INVESTIGATING COMPLEXITY IN TRANSCRIPTOME
EXPRESSION, REGULATION, AND EVOLUTION
USING MATHEMATICAL MODELING
By
Nicholas Louis Panchy

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Genetics – Doctor of Philosophy
2017

ABSTRACT
INVESTIGATING COMPLEXITY IN TRANSCRIPTOME EXPRESSION, REGULATION,
AND EVOLUTION USING MATHEMATICAL MODELING
By
Nicholas Louis Panchy
To date, gene expression has been characterized in over one thousand species across
more than a million experimental conditions. With this wealth of data, it is possible to investigate
the role that differential expression has in key biological processes, such as development, stress
response, and cell division. However, the complexity of the transcriptome makes the analysis of
expression challenging, as a single genome can contain thousands of genes as well as millions of
potential regulatory interactions shaped by more than a billion years of evolution. To address this
complexity, we can use the language of mathematics to create models of gene expression,
regulation, and evolution that define the system in a testable format. In the following chapters, I
will present research that applies mathematical modeling to the identification and regulation of
cyclically expressed genes as well as the evolution of transcriptional regulators following whole
genome duplication.
Cyclically expressed genes were studied in two systems. First, I investigated day-night
cycling or ‘diel’ genes in Chlamydomonas reinhardtii. Diel genes were identified de novo using
two models of cyclic expression that jointly classified half of all genes in C. reinhardtii as diel
expressed. To understand the regulation of diel expression, I clustered diel genes according their
peak of expression, or ‘phase’, and searched for cis-regulatory elements enriched (CREs) in the
promoters of each cluster. While I found putative CREs corresponding to each cluster, using
these CREs to predict diel expression using machine learning performed poorly compared to
previous models of expression regulation. Therefore, I changed systems to Saccharomyces

cerevisiae and studied cyclic expression during the cell cycle. Here, I applied machine learning
models to predict cell-cycle expression using regulatory interactions from four different data
sets. These models out performed the previous model of cyclic expression when using
regulatory interactions defined by chromatin-immunoprecipitation, transcription factor knockout
experiments, and position weight matrices. Further gains in performance were obtained by
combing interactions across data sets and using co-regulation by pairs of regulators involved in
feed-forward loops. The most important interactions for predicting cell-cycle expression
included not only known cell-cycle regulators but also two groups of transcription factors not
previously identified as being involved in cell-cycle regulation.
The evolution of transcriptional regulation was studied in Arabidopsis thaliana, which
has undergone several rounds of whole genome duplication (WGD), after which transcriptions
factors (TFs) are preferentially retained. Here, I applied maximum likelihood estimation to infer
the most likely ancestral expression and regulatory state of pairs of duplicate TFs prior to WGD.
Comparing this ancestral state to the existing TF duplicates, I found that one duplicate, the
“ancestral’ copy, tends to retain the majority of ancestral expression state and CREs, while the
other ‘non-ancestral’ copy loses ancestral expression and CREs, but also gains novel CREs
instead. Modeling the evolution of TFs pairs using as system of ordinary differential equations, I
demonstrated that the partitioning of ancestral states amongst duplicates is not random, but
occurs because the loss of ancestral expression occurs orders of magnitude faster in the first copy
than in the second. This suggests that TFs duplicate pairs are preferentially maintained such that
one copy is ‘ancestral’ and the other is not. Taken as a whole, the research in this dissertation
demonstrates how mathematical modeling can be applied to studying the expression, regulation
and evolution of the transcriptome.

Dedicated to…
…my family at home
…my family in the Shiu Lab
…my family from Providence
Thank you all for your love, support, and patience

iv

ACKNOWLEDGEMENTS

First, I would like to thank Dr. Shin-Han Shiu for agreeing to be my mentor. It was
talking with Shin-Han that helped convince me that MSU was where I wanted to do my graduate
studies and over the years his guidance, advice, and critique have been invaluable to my growth
as a scientist.
I would also like to thank my committee: Drs. David Arnosti, Eva Farre, and Chris
Adami for their input and feedback on my research. I also greatly appreciate the candor with
which they were willing to discuss the realities of being an academic as it stopped me from
becoming discouraged during my time as a student. Additionally, I would like to thank the
Genetics Graduate Program- particularly Barb Sears, Brian Schutte, Cathy Ernst, and Jeannine
Lee- as well as all the Genetics students I have shared my time with for helping navigate life as a
graduate student.
I owe a great deal to the past and current members of the Shiu Lab- Melissa, Gaurav,
Gunagxi, Sahra, Alex, Dave, Ming, Zing, Johnny, Beth, Christina, Peipei, Siobhan, Reid, and the
many visiting scholars and undergraduate students we worked with over the years. After all the
time we have spent together inside and outside of the lab, I feel that share a special sense of
camaraderie with you all. I am looking forward to hearing of your future successes even though I
may no longer be with you.
Finally, I would like to thank my parents for encouraging me to pursue my education and
supporting me through my years in college and graduate school and my family at Providence
PCA for providing me a home in Michigan.

v

TABLE OF CONTENTS

LIST OF TABLES……………………………………………………………………

ix

LIST OF FIGURES……………………………………………………………….......

xi

KEY TO ABBREVIATIONS………………………………………………………...

xiv

CHAPTER 1: INTRODUCTION……………………………………………………..
MOLECULAR MECHANISMS OF GENE REGULATION ……………….
IDENTIFYING REGULATORY INTERACTIONS…………………………
DEFINING GENE EXPRESSION…………….……………………………...
APPROACHES FOR MODELING GENE EXPRESSION………………….
APPLICATIONS FOR GENE EXPRESSION MODELS……………….......

1
3
4
6
10
13

CHAPTER 2: PREVALENCE, EVOLUTION, AND CIS-REGULATION OF DIEL
TRANSCRIPTION IN CHLAMYDOMONAS REINHARDTII………………………..
ABSTRACT…………………………………………………………………...
INTRODUCTION…………………………………………………………….
RESULTS AND DISCUSSION………………………………………………
Cycling gene expression is extensive in the C. reinhardtii genome…..
Phases of cycling gene expression are associated with a succession of
biological functions……………………………………………………
C. reinhardtii and A. thaliana orthologs have limited conservation in
cycling gene expression patterns ……………………………………...
Conservation of cyclic expression is more prevalent amongst older
duplicate genes…………………………………………………………
Cycling genes are enriched for specific putative Cis-regulatory elements
Phase of cyclic expression can be predicted for groups of genes with
common expression patterns or common function…………………….
CONCLUSIONS……………………………………………………………..
MATERIALS AND METHODS………………………………………………
Growth of Chlamydomonas reinhardtii Cultures………………………
RNA-sequencing……………………………………………………….
Identification of Cycling Genes………………………………………..
Clustering Cycling Genes According to Phase………………………...
Conservation of cyclic expression and phase of expression amongst
duplicate genes…………………………………………………………
Modeling cycling state divergence of duplicate genes…………………
Identification of putative cis-regulatory elements and phase prediction
Identifying groups of genes with common expression or common
function………………………………………………………………..
ACKNOWLEDGEMENTS…………………………………………………...
APPENDIX……………………………………………………………………
vi

17
18
19
22
22
24
28
32
36
40
44
47
47
48
48
50
51
51
52
53
54
55

CHAPTER 3: PREDICTING CELL-CYCLE EXPRESSED GENES IDENTIFIES
CANONICAL AND NON-CANONICAL REGULATORS OF TIME-SPECIFIC
EXPRESSION IN SACCHROMYCES CEREVISIAE……………………………........
ABSTRACT…………………………………………………………………..
INTRODUCTION…………………………………………………………….
RESULTS AND DISCUSSION………………………………………………
Comparing TF-target interactions from multiple regulatory data sets
Predicting timing of expression in the S. cerevisiae cell-cycle using
direct regulatory interactions……………………………………….....
Predicting timing of expression during the S. cerevisiae cell-cycle
using feed-forward loops………………………………………………
Using feature importance to merge GRNs and improve prediction of
cell-cycle expression…………………………………………………..
Functions of TFs important for predicting cell-cycle expression……..
Identifying regulatory modules for cell-cycle expression………….....
CONCLUSIONS………………………………………………………………
MATERIALS AND METHODS……………………………………………..
TF-target interaction data and regulatory cite mapping……………….
Overlap between TF-target interaction data…………………………..
Expected feed-forward loops in S. cerevisiae regulatory networks…..
Validating FFLs in cell-cycle expression……………………………..
Classifying cell-cycle genes using machine learning………………….
Evaluating the relationship between model performance, class and
feature…………………………………………………………………
Importance of features to predicting cell-cycle expression……………
GO Analysis…………………………………………………………..
ACKNOWLEDGEMENTS…………………………………………………..
APPENDIX……………………………………………………………………
CHAPTER 4: EXPRESSION AND REGULATORY ASYMMETRY IS A
FEATURE OF RETAINED TRANSCRIPTION FACTOR DUPLICATES………….
ABSTRACT…………………………………………………………………...
INTRODUCTION……………………………………………………………..
RESULTS AND DISCUSSION……………………………………………….
Retention of duplicate genes in different function groups following
WGD……………………………………………………………………
Linear model of WGD-duplicate retention across function groups…….
Features explaining degrees of retention across function groups and
WGD events……………………………………………………………
Partitioning of ancestral expression states following TF duplication….
Influence of the timing of TF duplication and expression state
evolution……………………………………………………………….
Asymmetry in the partitioning of ancestral expression and regulatory
sites……………………………………………………………………
Patterns of WGD-duplicate divergences and partitioning results from
vii

88
89
90
95
95
101
105
111
116
119
128
130
130
130
131
131
132
133
134
134
135
136
175
176
177
180
180
184
187
191
195
197

evolutionary bias………………………………………………………
CONCLUSIONS………………………………………………………………
MATERIALS AND METHODS……………………………………………...
Genome sequences, gene annotation, and Expression Data…………...
Defining TFs and other groups of genes in A. thaliana………………..
Fitting odds ratio of duplicate retention within each group of genes for
each WGD event using linear models………………………………….
Inferring ancestral expression levels and cis-regulatory sites………….
Asymmetry of the retention of ancestral expression and regulatory
sites…………………………………………………………………….
Ordinary differential equation models of TF state evolution…………..
ACKNOWLEDGEMENTS……………………………………………………
APPENDIX…………………………………………………………………….

204
209
212
212
213

CHAPTER 5: CONCLUSION…………………………………………………………
PREDICTING CYCLIC EXPRESSION PATTERNS USING
CIS-REGULATORY ELEMENTS……………………………………………
DUPLICATION AND EVOLUTION OF TRANSCRIPTION FACTORS …
FUTURE PROSPECTS FOR MODELING GENE EXPRESSION………….

275

REFERENCES…………………………………………………………………………

282

viii

214
215
216
218
220
221

276
278
280

LIST OF TABLES

Supplemental Table 2.1 Distribution of Fourier Transform cyclic score and
COSPOT p-values………………………………………………………………………

70

Supplemental Table 2.2 Descriptions of the GO terms in each of the five broad
functional categories……………………………………………………………………

72

Supplemental Table 2.3 Optimal parameters and performance measures of SVM
classification……………………………………………………………………………

75

Supplemental Table 2.4 “Gold Standard” cycling genes in C. reinhardtii……………..

76

Supplemental Table 2.5 Performance COSPOT and DFT on C. reinhardtii…………..

78

Supplemental Table 2.6 Performance of combining COPSOT and DFT on C.
reinhardtii………………………………………………………………………………

79

Table 3.1 Size and origin of GRNs defined from each data set……………………….

96

Table 3.2 Observed and expected number of FFLs in GRNs defined using different
data sets………………………………………………………………………………..

109

Supplemental Table 3.1 Coverage of cell-cycle genes by TF-target interactions in each
data set…………………………………………………………………………………

149

Supplemental Table 3.2 Coverage of cell-cycle genes by FFL interactions in each data
set………………………………………………………………………………………

150

Supplemental Table 3.3 Performance of classifiers built using TF-target interactions on
only cell-cycle genes covered by ChIP-Chip FFLs……………………………………

151

Supplemental Table 3.4 Total number of feature present in each model built from
combined features sets…………………………………………………………………

152

Supplemental Table 3.5 Enrichment of TFs with cell-cycle regulation GO annotation in
features of the ChIP-Chip and Deletion data sets………………………………………

153

Supplemental Table 3.6 Over and under enrichment of GO Terms in ChIP-Chip and
Deletion feature sets……………………………………………………………………

154

Supplemental Table 3.7 Over enrichment of GO Terms in ChIP-Chip and Deletion
feature sets for specific phases of cell cycle expression ………………………………

163

ix

Table 4.1 Performance of best fitting models of the odds ratio of duplicate retention…

188

Table 4.2 Importance of features used in the linear models of duplicate retention…….

189

Supplemental Table 4.1 Data sets used in linear model of duplicate retention………..

235

Supplemental Table 4.2 Subsets of AtGenExpress used for ancestral expression
inference………………………………………………………………………………..

239

Supplemental Table 4.3 Observed and expected frequency of duplicates TF pairs in
a conserved, partitioned, and diverged state……………………………………………

240

Supplemental Table 4.4 Experimental conditions used in each subset of
AtGenExpress…………………………………………………………………………..

241

Supplemental Table 4.5 RNA-seq data sets……………………………………………

242

Supplemental Table 4.6 Genes belonging to each GO-defined function group………..

243

Supplemental Table 4.7 TF genes belonging to each TF family in A. thaliana………..

266

Supplemental Table 4.8 Best fit parameters of ODE models of the evolution of TF
expression above or below the ancestral state………………………………………….

272

Supplemental Table 4.9 Best fit parameters of ODE models of partitioning of
ancestral states between duplicate TFs…………………………………………………

273

Supplemental Table 4.10. The importance of all features used in the classification of
individual duplicate genes………………………………………………………………

274

x

LIST OF FIGURES

Figure 2.1 Period, amplitude, and phase of cyclic expression…………………………

23

Figure 2.2 Phase of gene expression and cyclically expressed GO terms……………..

25

Figure 2.3 Phase specific expression of broad functional categories………………….

29

Figure 2.4 Conservation of cyclic expression and phase of cyclic expression………...

33

Figure 2.5 Top three pCREs enriched in each phase cluster of cyclic genes………….

38

Figure 2.6 Enrichment and performance of phase-specific pCREs……………….. ….

39

Figure 2.7 Expression of best predicted co-expression cluster and GO terms…….. ….

42

Supplemental Figure 2.1 Period, amplitude, and phase of cyclic expression amongst
predictions made by COSPOT, DFT, and both methods combined……………………

59

Supplemental Figure 2.2 Most over- and under-enriched GO terms amongst phase
clusters of cycling genes………………………………………………………………...

61

Supplemental Figure 2.3 Divergence of duplicate gene expression state modeled as a
system of difference equations………………………………………………..................

63

Supplemental Figure 2.4 Precision-recall and AUC-ROC curves of SVM predictions
for C. reinhardtii………………………………………………………………………..

64

Supplemental Figure 2.5 Regression of the AUC-ROC of phase-expression clusters
against cluster size, and Pearson Correlation Coefficient (PCC) of genes in the cluster..

66

Supplemental Figure 2.6 Expression profiles of cell cycle genes (MAT3, E2F,
CDKA1, and CDKB1) in C. reinhardtii grown in TAP (Tris-Acetate-Phosphate)
culture…………………………………………………………………………………..

68

Supplemental Figure 2.7 Distribution of Fourier Transform cyclic score and COSPOT
p-values………………………………………………………………………………...

69

Figure 3.1 Coverage of TF and TF-interactions by data set……………………………

97

Figure 3.2 Overlap in TF-target interactions across data sets………………………….

100

Figure 3.3 Performances of classifiers using TF-target interactions across all data sets

103

xi

Figure 3.4 Performance of classifiers using only FFLs across all data sets …………

106

Figure 3.5 Performance of classifiers built using important features from ChIP-Chip,
Deletion, and combined ChIP-Chip/Deletion data set…………………………………

113

Figure 3.6 The cell-cycle expression GRN defined using the 10th percentile of ChIPChip Features…………………………………………………………………………...

121

Figure 3.7 The cell-cycle expression GRN defined using the 25th percentile of
ChIP-Chip TF-TF interactions…………………………………………………………

125

Supplemental Figure 3.1 Expected overlaps of TF-target interactions across regulatory
data sets………………………………………………………………………………...

137

Supplemental Figure 3.2 Expression profiles of genes expressed at specific phases of
the cell-cycle……………………………………………………………………………

138

Supplemental Figure 3.3 Performance of classifier using alternative feature sets…….

140

Supplemental Figure 3.4 Relationship between TF genes and TF-TF interactions……

142

Supplemental Figure 3.5 Overlap of FFLs across data sets……………………………

144

Supplemental Figure 3.6 Importance of TF features across classification models…….

145

Supplemental Figure 3.7 The cell-cycle expression GRN defined using the 25th
percentile of Deletion TF-TF interactions……………………………………………..

147

Figure 4.1 Retention of WGD-duplicate genes in A. thaliana…………………………

182

Figure 4.2 Linear model of the degree of duplicate retention in function groups based
on genes features……………………………………………………………………….

185

Figure 4.3 Evolution of expression in TF WGD-duplicates……………………………

193

Figure 4.4 Asymmetry of ancestral state retention in TF WGD-duplicates……………

198

Figure 4.5 Expression partitioning between duplicate pairs with high regulatory
asymmetry………………………………………………………………………………

202

Figure 4.6 ODE models of TF WGD-duplicate expression and cis-regulatory site
evolution relative to the ancestral state…………………………………………………

206

Supplemental Figure 4.1 Frequency distribution of synonymous substitution rate (Ks)
between putative paralogs………………………………………………………………

227

xii

Supplemental Figure 4.2 Difference between the observed rates of duplicate retention
and the rates predicted by the linear models of duplicate retention…………………….

228

Supplemental Figure 4.3 Difference in expression quartile of individual TF duplicates
compared to their ancestral state……………………………………………………….

229

Supplemental Figure 4.4 Deviation of pairs of TF WGD-duplicates from their
ancestral state………………………………………………………………………….

230

Supplemental Figure 4.5 ODE models of the evolution of ancestral expression into
either a higher or lower expression quartile……………………………………………

231

Supplemental Figure 4.6 ODE models of TF WGD-duplicate expression evolution
relative to ancestral state for the Ctrl, Diff, and Stress expression subsets…………….

233

xiii

KEY TO ABBREVIATIONS

ANOVA

Analysis of variance

Asy

Asymmetry score

AUC-ROC

Area under the curve of the receiver operating characteristic

BLAST

Basic alignment search tool

bp

base pair

ChIP

Chromatin immunoprecipitation

CPM

Counts per million

CRE

Cis-regulatory element

Ctrl

Control

DAP-Seq

DNA affinity purification sequencing

Diel

24-hour, day night period

DNA

Deoxyribonucleic acid

DFT

Discrete Fourier transform

Diff

Differential Expression

FFL

Feed –forward loop

FPKM

Fragments per kilo-base of transcript per million mapped reads

G1

Initial growth (cell-cycle)

G2

Intermediate growth (cell-cycle)

GEO

Gene Expression Omnibus

GO

Gene ontology

GRN

Gene regulatory network

xiv

Ka

Non-synonymous substitution rate

Ks

Synonymous substitution rate

Kb

kilo-basepair

IQR

Inter-quartile range

LightDev

Light and Development

miRNA

micro RNA

M

Cell division (cell-cycle)

MAD

Mean absolute deviation

MSU

Michigan State University

NCBI

National Center for Biotechnology Information

ODE

Ordinary differential equation

PBM

Protein binding microarray

PCC

Pearson’s correlation coefficient

PCR

Polymerase chain reaction

pCRE

putative CRE

PHYLIP

Phylogeny Inference Package

PWM

Position weight matrix

pv

P-value

r2

Coefficient of determination

RAxML

Randomized Axelerated Maximum Likelihood

RIN

RNA Integrity Number

RMA

Robust Multichip Average

RNA

Ribonucleic acid

xv

RNA-Seq

RNA sequencing

RPKM

Reads per kilo-base of transcript per million mapped reads

S

DNA replication (cell-cycle)

SRA

Sequence Read Archive

SVM

Support Vector Machine

TAP

Tris-Acetate-Phosphate

TF

Transcription factor

TPM

Transcripts per million

WEKA

Waikato Environment for Knowledge Analysis

WGD

Whole genome duplication

ZT

Zeitgeber time

xvi

CHAPTER 1: INTRODUCTION

1

Following the advent of microarray (Schulze and Downward, 2001; Hoheisel, 2006) and
high-throughput sequencing (Reuter et al., 2015; Goodwin et al., 2016) technology, gene
expression has been inferred using transcript quantification in over 3300 species, with more than
400 having in excess of 100 samples publically available through online databases (GEO,
https://www.ncbi.nlm.nih.gov/geo/summary/?type=taxfull). With such a breadth of expression
data available, in terms of transcriptome coverage, organisms, and conditions, it has become
possible to characterize genes using their expression profiles. The analysis of these profiles has
been applied to a variety of research questions: the progression and outcome human disease
(Henriksen and Kotelevtsev, 2002; van ’t Veer et al., 2002; Bergholdt et al., 2009; CooperKnock et al., 2012), understanding the basis for variation in quantitative phenotypes (JiménezGómez et al., 2010; Jimenez-Gomez et al., 2011; Nica and Dermitzakis, 2013; Albert and
Kruglyak, 2015), and predicting the phenotype/functions of genes (van Noort et al., 2003; Takagi
et al., 2014; Lloyd et al., 2015; Uygun et al., 2016). Yet, while expression profile analysis can be
useful for identifying and classifying genes, the question remains as to how patterns of
expression are established and maintained. One approach to understanding how expression
patterns are regulated is the use of mathematical modeling: the representation of a system using
mathematical objects (variables, operators, equations, etc). For gene expression in particular, this
involves defining using set of explanatory variables to predict the expression of genes as
accurately as possible in order to answer a biological question about the genes, their regulators,
or the dynamics of the system. Although the molecular mechanisms that regulate gene
expression are understood (Lee and Young, 2000; Lelli et al., 2012; Voss and Hager, 2014) and
broad patterns of expression can be inferred from sequence alone (Beer and Tavazoie, 2004),
modeling gene expression remains a challenging task, particularly in response to specific

2

environmental conditions (Zou et al., 2011), cellular location (Uygun et al., 2017), and time
(Panchy et al., 2014). My research focuses on the application of differential equations and
machine learning models to understanding the regulation of cyclic expression and evolution of
regulatory systems, but many different modeling approaches have been applied to an equally
varied set of biological questions.
MOLECULAR MECHANISMS OF GENE REGULATION
As gene expression is often quantified by the level of mRNA transcript, approaches to
modeling gene expression are guided by what is known about the regulation of transcription at
the molecular level. The transcription of a gene is primarily (but not exclusively) regulated at the
initiation stage when the RNA-polymerase complex is recruited to the promoter region upstream
of the transcription start site (Lee and Young, 2000). This promoter region contains a core
promoter element, which is ubiquitous in function across eukaryotic genes, that binds the
components of the RNA-polymerase complex, which is also common across eukaryotes.
However, the core promoter alone is insufficient for regulation of transcription in vivo, and
additional factors, called transcription factors (TFs), are required to enhance or repress RNApolymerase binding and activity (Lee and Young, 2000). Because modeling is primarily focused
on differences in expression either between genes or in a single gene across time or a set of
conditions, the common core elements can be ignored in favor of the activity of TFs.
The affinity of TFs for a particular promoter primarily depends on regions of DNA
known as cis-regulatory elements (CREs), such that the presence or absence of these elements
represent TF regulation. However, there is not a 1-to-1 relationship between TFs and CREs, but
rather a single TF can bind multiple CREs with varying degrees of sequence similarity, and a
single base change may or may not disrupt the binding potential of an element depending on the

3

position of the change (Badis et al., 2009). Furthermore, whether or not a TF interacts with a
promoter also depends on chromatin state, nucleosome positioning, histone modification, and
cooperativity with other TFs (Lee and Young, 2000; Lelli et al., 2012; Voss and Hager, 2014).
Although the presence or absence of CREs is relatively fixed in a given genome, the accessibility
of chromatin, histone-code and concentration of other TFs are all dynamic, meaning that CREs
alone are often insufficient to determine if a TF regulates a gene under a specific set of
conditions and thus plays a role in regulation. Therefore, the first step in modeling gene
expression is determining how to identify a set of relevant regulatory interactions.
IDENTIFYING REGULATORY INTERACTIONS
The types of evidence that can be used to identify regulatory interactions can be divided
into three broad categories: directly assaying protein-DNA interactions, prediction of TF binding
from promoter sequence, and inferring interaction between genes based on expression variation.
The DNA sequence(s) that a protein will bind to can be assayed either in vitro using protein
binding microarrays (Bulyk, 2007; Berger and Bulyk, 2009), or in vivo with chromatin
immunoprecipitation (Buck and Lieb, 2004; Furey, 2012) or DNA affinity purification
(O’Malley et al., 2016). Because all approaches used to define regulation are based on binding
sequences, it is necessary for a genome to be sequenced and annotated so that the recovered
sequence can be mapped to the promoter region of potential target genes. Often, post-processing
is required to address noisy TF binding (i.e. false-positive interactions). For high-throughput
sequencing approaches in particular (e.g. ChIP-seq and DAP-seq), high-confidence binding sites
can be identified by mapping sequences reads to the genome and using software (Zhang et al.,
2008) to call “peaks” where multiple-reads overlap the same sequence (Feng et al., 2012; Landt
et al., 2012; O’Malley et al., 2016). However, experimental evidence may not be available for

4

every TF in an organism and, given that there are dozens of sequenced eukaryotic genomes with
more than 1000 TF genes (Kummerfeld and Teichmann, 2006), assaying binding of the entire set
of transcription factors may not be feasible in some cases.
In the absence of direct binding information for TFs, the presence of “putative” cisregulatory elements (Beer and Tavazoie, 2004; Zou et al., 2011; Liu et al., 2015; Uygun et al.,
2017) may be inferred from the promoter sequence of target genes (reviewed in Das and Dai,
2007 and Li et al., 2015). By using computational approaches for binding site prediction
(software such as AlignAce, YMF, MEME), it is possible to identify pCREs using promoter
sequences alone (Hughes et al., 2000; Sinha and Tompa, 2003; Bailey et al., 2006; Bailey et al.,
2009). There are also machine learning and deep learning methods for integrating multiple types
of omics data, including DNA accessibility, chromatin structure, histone marks, and available
binding site information from assays like ChIP (Pique-Regi et al., 2011; Hoffman et al., 2012;
Alipanahi et al., 2015; Li et al., 2016). Furthermore, transcription initiation events associated
with cis-regulatory elements can produce non-coding RNAs that can be captured by Global RunOn sequencing and used to infer pCREs from data (Danko et al., 2015). Predictive methods do
not necessarily connect pCREs to the TFs that bind them, but pCREs derived from the promoters
of co-expressed genes do show similarity to known TF binding motifs (Uygun et al., 2017),
which suggests that presence of pCREs do in fact reflect the binding potential of TFs. However,
whether assayed or predicted, TF-target gene interactions identified based on TF binding suffer
from the same drawback: binding potential does not guarantee regulatory function as actual TF
binding can be greatly affected by small changes in sequence (Kwasnieski et al., 2012) and the
presence of other TFs at the promoter site (Spivak and Stormo, 2016).

5

Alternatively, the interaction between TFs and target genes can be inferred based on
changes in gene expression. In this case, the presence of an interaction is assumed if two genes
share a “coordinated” pattern of expression, though what constitutes “coordinated” varies with
approach. Coordination of expression has been characterized using mutual information
(Margolin et al., 2006; Faith et al., 2007), regression (Geeven et al., 2012), differential equations
(Honkela et al., 2010), and Bayesian networks (Friedman et al., 2000). However, the results of
the Dialogue on Reverse Engineering Assessment and Methods (DREAM) network inference
challenge, an open challenge to infer gene regulatory networks from a standard set of synthetic
and actual expression data (Marbach et al., 2010; Marbach et al., 2012), suggest that ensemble
methods that combine multiple approaches have the best performance when predicting an
artificially generated gene regulatory network with 195 regulators and 1643 genes. However,
even the best ensemble methods perform poorly when applied to Escherichia coli (296
regulators, 4297 genes) and no better than random guessing on Saccharomyces cerevisiae (183
regulators, 5677 genes), suggesting that the performance of expression based methods decline as
the size of the network increases (Marbach et al., 2012). Furthermore, methods that predict
interactions based on expression tend to exhibit common errors, such as inferring relationships
between co-regulated genes where none exist (fan-out error), inflating the number of interactions
possessed by highly connected target genes (fan-in error), and inferring “shortcuts” between the
beginning and end of pathways (cascade error)(Marbach et al., 2010). Including expression from
TF-knockout experiments helps reduce fan-in and fan-out error, and TF-knockout data has been
used to directly infer interactions (Reimand et al., 2010). However, TF-knockout data was not
useful for addressing cascade errors in a model context (Marbach et al., 2012). In S. cerevisiae,
most TF-target interactions derived from TF-knockout studies lacked evidence of direct, in vivo

6

binding when compared with ChIP-Chip binding data, though there was often was evidence of
interaction through an intermediate TF. Keeping these caveats in mind, interactions inferred from
expression data can provide useful information for modeling expression, and results presented
later suggest that combing both binding data and TF-knockouts improves predictions of
expression. Yet interactions are only half of the equation: before a mathematical model of
expression can be made, what the model is trying to predict must also be defined.
DEFINING GENE EXPRESSION
Compared to identifying regulatory interactions, defining what the model is trying to
predict may seem trivial as the question often comes before the model. Yet, even if the set of
target genes and the pattern of interest are known beforehand, it is still be necessary to decide
how to define gene expression. At its most basic, modeling expression can be treated as a
quantification problem or a classification problem. Using stress response as an example,
quantification would involve comparing two continuous expression values before and after
stress, while classification would involve categorizing genes as up-regulated, down-regulated or
unchanged following stress. There is no “best” way to treat expression in this regard, but rather
how expression is defined should be guided by the question at hand; is it important to know that
gene expression changes or the magnitude of the expression changes that occur? Making this
distinction is an important first step to determining what type of data is needed, how to treat that
data, and what modeling approach to use.
Even if expression is treated as a classification problem, categorizing or identifying
expression patterns often begins with quantifying the amount of mRNA transcript in a sample.
Several technologies are currently available to quantify transcript levels, including Northern
blotting (Fernyhough, 2001), fluorescence in situ hybridization (Femino et al., 1998), reverse

7

transcription PCR (Bustin, 2000; Nolan et al., 2006), microarrays (Xiang and Chen, 2000) and
high-throughput sequencing (Wang et al., 2009). Of the more than two million expression data
sets publically available through GEO, 96.6% are derived from microarray or sequencing, with
PCR a distant third (1.3%). Microarray data from GEO can be accessed and analyzed using
BioConductor (Gentleman et al., 2004; Davis and Meltzer, 2007) while sequencing data needs to
pre-processed, mapped, and quantified (reviewed in Conesa et al., 2016). Metrics for quantifying
sequenced reads comes in two types: (1) counts/transcripts per million (CPM/TPM), in which the
number of reads/assembled transcripts is adjusted based on the total number of mapped
reads/transcripts in millions, and (2) reads/fragments per kilobase of transcripts per million
mapped reads (RPKM/FPKM). In general, RPKM/FPKM is preferred for comparing expression
within samples because the longer transcripts tend to produce more reads, while CPM/TPM
preferred for comparing across samples/species (Conesa et al., 2016).
Expression can also be quantified as the difference in expression between genes across
samples (e.g. treatment vs. control). For microarray data, BioConductor provides a protocol for
differential expression (see https:// www.bioconductor.org/help/workflows/arrays/), but choosing
the best approach for sequencing data depends on the number of experimental conditions,
number of replicates per condition, sample size and available computational resources (see
Soneson and Delorenzi, 2013 and Rapaport et al., 2013). Differential expression can also be
applied to classification problems. In this case, the significance and direction of differential
change can be used to classify genes as up-regulated, down-regulated, or not changed under
specific treatments, though it is not uncommon to require a minimum level of change relative to
control conditions as well (Kilian et al., 2007; Wu et al., 2015; Uygun et al., 2017). In the case of
multiple treatment conditions, this scheme can be applied to each condition independently, but

8

the number of possible classes will increase quickly (3N) and require multiple-hypothesis testing.
Software like edgeR (Robinson et al., 2010) and DESeq (Anders and Huber, 2010) can be used
to directly test the significance of defined patterns of differential and non-differential expression
across multiple conditions; however, if specific patterns of expression are not known, clustering
can be applied to identify patterns of expression de-novo (Kerr et al., 2008; Oyelade et al., 2016).
Differential expression is not the only criteria for classifying genes by expression. In the
case of long time series, progressive or repeated change in expression relative to the mean or
starting level of expression may be of interest. A good example of this is cyclic patterns of gene
expression, such as occurs across the cell-cycle (Spellman et al., 1998) or in response to the
circadian rhythm (Chen et al., 1998; Sukumaran et al., 2010). Approaches to identifying cyclic
expression have employed both models of cyclic expression (Straume, 2004; Hughes et al.,
2010) and the underlying periodicity expected of cyclic expression (Wichert et al., 2004; Panchy
et al., 2014). Although these models are specific to cyclic expression, the same sort of approach
can be applied to any pattern of expression.
A final consideration for classifying genes by expression is how to define a negative set,
i.e. a set of genes without the desired pattern. Often, this is not as simple as using all other genes
because genes which lack the target pattern of expression are not all alike. Therefore, it can be
advantageous to define a negative set of genes using its own, separate pattern of expression. For
example, when classifying salt-responsive genes, Zou et al. used negative genes that were not
differentially expressed under any stress because of possible cross-talk between genes expressed
under different stress conditions (Zou et al., 2011). The decision of how to define a negative set
will also be influenced by what approach is used to model expression because certain methods,
such as machine learning, are more sensitive to the choice of negative examples. Ideally, the

9

overall process of defining expression will be co-simultaneous with model development in order
to avoid conflict.
APPROACHES FOR MODELING GENE EXPRESSION
Defining the modeling problem will also influence the approach used to model gene
expression. At its most basic, the expression of a gene can be discretized as either active (1) or
inactive (0), allowing interactions between genes and their regulators to be defined using logical
operations (i.e. AND, OR, NOT) (Karlebach and Shamir, 2008; Ay and Arnosti, 2011). In this
form, expression can be modeled using Boolean networks (Glass and Kauffman, 1973; Thomas,
1973; Kauffman et al., 2003). Boolean networks have been used to study the robustness and
stability of GRNs in a variety of systems including S. cerevisiae (Kauffman et al., 2003), human
cancers (Shmulevich et al., 2003; Trairatphisan et al., 2016), and Drosophila melanogaster
(SĂĄnchez and Thieffry, 2001; Yuh et al., 2001; Albert and Othmer, 2003). This method can also
be extended to cases where there is imperfect information about regulatory interactions by using
probabilistic Boolean networks (Shmulevich et al., 2002; Shmulevich et al., 2003). However,
Boolean networks fail to accurately capture the behavior of certain biological interactions,
particularly cases where a gene negatively regulates its own expression (Karlebach and Shamir,
2008; Ay and Arnosti, 2011).
Another issue is that analyzing Boolean networks becomes increasingly difficult with
increasing size of the GRN being studied (Karlebach and Shamir, 2008). The number of possible
global states in a network grows according to the number of states per gene (k) and the number
of genes (n) in exponential fashion (kn). Therefore, the number of global states in even a
relatively small genome such E. coli K-12 becomes prohibitively large (24500 ~ 101350). For this
reason, it is often beneficial to cluster co-expressed genes together so that, instead of modeling

10

the global pattern of expression across all genes, the problem becomes correctly assigning genes
to a finite number of co-expression modules. This clustering approach was taken by Beer and
Tavazoie (Beer and Tavazoie, 2004), who used Bayesian networks to assign S. cerevisiae genes
to one of 49 representative clusters, and was extended by Yuan et al. (2007) with the use of a
naive Bayes classifier. Though Beer and Tavazoie (2004) used k-means clustering in order to
construct gene expression modules, other clustering methods are available, including hierarchical
clustering, self-organizing maps, self-organizing tree algorithms (Yin et al., 2006), as well as
more than a dozen different distance metrics (Jaskowiak et al., 2014).
Alternatively, classification, either in the form of discretization or clustering, can be
avoided altogether and the quantitative measures of expression taken from experimental data can
be modeled directly. Using linear models, gene expression can be modeled from only expression
data by assuming each regulator functions independently and its net effect on the target gene is
summarized by a singular weight value. However, while linear models have been used to infer
regulatory interactions (Yeung et al., 2002; Bansal et al., 2006) and understand risk factors in
human disease (Li et al., 2014; Trabzuni et al., 2014), they cannot applied to questions about the
dynamics of molecular regulation because the behavior of this system is non-linear (Karlebach
and Shamir, 2008). In contrast, thermodynamic models (Bintu et al., 2005; Segal et al., 2008)
and Michaelis–Menten kinetics (Nachman et al., 2004) have been used to account for the
concentration-dependent nature of TF binding to CREs using probabilistic binding and nonlinear functions, respectively. Notably, the Michaelis-Menten equation was derived as a solution
to a system of ordinary differential equations (ODEs) describing enzyme kinetics under certain
assumptions (Schnell, 2014; Wong et al., 2015). Other systems of ODEs have been used to
incorporate different assumptions and variables into models of expression such variable cell

11

mass and volume (Chen et al., 2004; Li et al., 2008), spatial context and diffusion (Eldar et al.,
2002; Jaeger et al., 2004), and separate binding mechanics for protein regulators and microRNAs
(Zhang et al. 2014; Hong et al. 2015). However, both systems of ODEs and thermodynamic
models are sensitive to the choice of regulatory interactions, such that the erroneous omission or
addition of a single regulator can potentially have a significant effect on the outcome (Ay and
Arnosti, 2011).
Though they are most obvious in complex models of quantitative expression measures,
all modeling approaches described so far make assumptions about how regulators function to
control the expression of their targets. Alternatively, the problem of modeling gene expression
can be approached by trying to “learn” what features are important for regulating expression
using machine learning algorithms (reviewed in Libbrecht and Noble, 2015). Rather than create
an explicit model of how gene expression is regulated, these approaches employ programs
designed to optimize some task (in this case, the prediction of gene expression) from a set of
features (regulatory interactions and any other data). This approach represents a double edged
sword in that machine learning algorithms can incorporate many different types of data without
prior knowledge of how they function in a system and assess their importance to controlling gene
expression, but little can be interpreted about why a specific feature is important from the
resulting model. Traditional machine learning algorithms, such as support vector machines and
random forest, have been applied to understand the effects of combinatorial regulation (Zou et
al., 2011), nucleosome positioning (Liu et al., 2015), and tissue-specific regulation (Uygun et al.,
2017) on gene expression in Arabidopsis thaliana as well as the influence chromatin state on
general expression in Caenorhabditis elegans (Cheng et al., 2011) and human cell lines (Cheng
et al., 2011; Dong et al., 2012). Furthermore, so called “deep learning”, which uses multi-layered

12

neural networks, has recently been applied to predict gene expression using expression data
(Chen et al., 2016b) and histone modification (Singh et al., 2016). This method in particular
holds great promise for biological research, not only because it has the potential to outperform
traditional machine learning methods (Singh et al., 2016), but also because there have recent,
rapid advances in this technology (Chen et al., 2016a; Min et al., 2016; Silver et al., 2016;
Fernando et al. 2017) that promise new opportunities for applying deep learning to the biological
sciences.
APPLICATIONS FOR GENE EXPRESSION MODELS
Ultimately, the objective of all expression models is to accurately predict expression in
the target set of genes, and this predictive function alone is sufficient to answer biological
questions. The resulting models can also be used to explore the dynamics of the GRN being
modeled as well as discover new elements important for regulating expression. An example of a
direct application of predictive models includes Li et al. (2008) who built an ODE model of cell
division (including the expression of key regulatory and structural genes) in the stalked cells of
Caulobacter crescentus. The parameters of the model were fit using the expression values in
wild-type cells and subsequently validated by testing if known mutant phenotypes mutants could
be reproduced by modifying the network to mimic the mutation. Expect when a mutation
involved a process outside of the model (e.g. phosphorylation of regulators), the Li model was
able to recreate mutant phenotypes and was subsequently used to predict the phenotype of
previously uncharacterized mutants. Similarly, Chen et al. (2004) constructed an ODE model of
the cell cycle of S. cerevisiae that could accurately model the wild-type cell-cycle as well as the
phenotypes of 92% of characterized mutants. In some cases, predictions made about novel
mutants were independently validated by another research group (Archambault et al., 2003).

13

However, Chen et al. noted that, though their model robustly predicted expression during the
wild-type cell-cycle, accurately predicting mutant phenotypes was more sensitive to small
changes in parameter values (2004). Hence, because of this sensitivity, it is reasonable to treat
novel phenotype predictions with skepticism even when the underlying model accurately
characterizes expression under normal conditions.
Predictions are not the sole purview of expression models, and often it is the model itself
which is of interest, as it can be used to explore the dynamics of the system. Li et al. (2004)
constructed a Boolean network of 11 key regulators of the S. cerevisiae cell-cycle. They found
that all possible initial conditions eventually progressed into one of seven steady states, with
most (86%) initial conditions resulting in a steady state representative of the G1 phase, the
resting state of the cell cycle. Furthermore, artificially inducing the cell-cycle (i.e. activating
Cln3) in the model resulted in an unstable G1-phase state that evolved into an S-phase (DNAreplication)-like state, followed by G2 (intermediate growth), and M (cell division) before
returning the stable G1-phase state, mirroring normal progression through the cell cycle.
Importantly, perturbing the Boolean network by deleting or adding interactions most often did
not affect either the stability of the G1 state or the frequency with which other global states
evolved into the G1 state. This suggests that the robustness of cell-cycle progression is in part
due to the structure of its regulatory network. Another example of using expression models to
explore model dynamics is an ODE model of epithelial to mesenchymal transition (EMT) in
human cells lines (Hong et al., 2015). In addition to predicting the known reversibility of the
transition between epithelial and mesenchymal cell populations, the model also predicts the
existence of two stable intermediate states where cells express markers of both epithelial and
mesenchymal cells. By perturbing regulatory interactions in the model, Hong et al. found that the

14

stability of these intermediate cell types depends on feedback loops between transcription factors
(Ovol2 and Zeb1) and between miRNAs and transcription factors (miR34a and Snail1, miR200
and Zeb1). These intermediate states are of particular interest as certain human cancers display
characteristics of both epithelial and mesenchymal cells (Hong et al., 2015). In general, dynamic
models like these offer the advantage of being able to perturb complex systems in silico to guide
or supplement experimental approaches.
Finally, expression models have been used to discover important features of gene
regulatory systems by looking at differences in performance after including/excluding different
features. Although not the sole focus of their study, in building thermodynamic models of genes
which regulate segmentation in Drosophila, Segal et al. (2008) found that including CREs that
were neither enriched amongst segmentation gene promoters nor expected to bind to high
affinity transcription factors were nevertheless important to accurately predict expression. These
weak binding sites were found to be clustered with other cis-regulatory sites that bind the same
transcription factors, suggesting that they might play a role in cooperative binding, which is
important for predicting the sharp boundaries of expression between segments that are observed
in nature. Taking a different approach, Zou et al. (2011) used support vector machine to predict
genes in A. thaliana that are differentially expressed in response to stress based on the presence
of CREs in the promoter of the gene. In addition to experimentally identified CREs, Zou et al.
included computationally predicted pCREs enriched in the promoters of abiotic and biotic stressresponsive genes, respectively. Including these pCREs improved the performance of the model,
suggesting that they represent bona-fide binding sites for as of yet unidentified TFs. They also
identified pairs of CREs enriched amongst stress-responsive genes, the inclusion of which

15

further strengthened the prediction of stress response. Like the results of Segal et al., this finding
indicates that cooperative binding plays an important role in stress regulation.
In the following chapters, I will present three applications of modeling gene expression I
employed in my research. First, I used two models of cycling expression to identify diel
expressed genes in the green alga Chlamydomonas reinhardtii and cluster them according to the
timing of peak expression or phase. pCREs enriched in each phase-cluster were identified and
subsequently used to predict the expression phase of diel genes. In the next chapter, I further
explored predicting cyclic expression by comparing the performance of four different sets of
regulatory interactions defined based on experimental evidence in predicting cell-cycle
expression in S. cerevisiae. I also looked how the prediction performance was affected by
including network motifs such as feed-forward loops as features and combining the best features
from multiple data sets. Known cell-cycle regulators were identified as being amongst the most
important TFs for correctly predicting cell-cycle expression. However, interactions amongst TFs
that were neither individually important nor annotated cell-cycle regulators were also necessary
to accurately predict expression. In the final chapter, I describe a different approach to
understanding expression regulation, by modeling the evolution of ancestral expression and
regulatory states in duplicate pairs of TFs. A system of ODEs was used to model the loss of
expression and regulatory sites between these duplicate TFs, and this model suggests that
asymmetry between copies, were one duplicate retains ancestral states and other diverges, is
favored. Together these studies illustrate how expression modeling can be applied to a wide
variety of biological questions as well as answer questions about how cyclically expressed genes
are regulated and how the GRNs that control such complex patterns of expression may have
evolved.

16

CHAPTER 2: PREVALENCE, EVOLUTION, AND CIS-REGULATION OF DIEL
TRANSCRIPTION IN CHLAMYDOMONAS REINHARDTII1

1

The work described in this chapter has been published in the following manuscript

Nicholas Panchy, Guangxi Wu, Linsey Newton, Chia-Hong Tsai, Jin Chen, Christoph Benning,
Eva M. Farre, Shin-Han Shiu (2014) Prevalence, Evolution, and cis-Regulation of Diel
Transcription in Chlamydomonas reinhardtii. G3 4:2461-2471

17

ABSTRACT
Endogenous (circadian) and exogenous (e.g. diel) biological rhythms are a prominent
feature of many living systems. In green algal species, knowledge of the extent of diel
rhythmicity of genome wide gene expression, its evolution, and its cis-regulatory mechanism is
limited. In this study, we identified cyclically expressed genes under diel conditions in
Chlamydomonas reinhardtii and found that ~50% of the 17,114 annotated genes exhibited cyclic
expression. These cyclic expression patterns indicate a clear succession of biological processes
during the course of a day. Among 237 functional categories enriched in cyclically expressed
genes, >90% were phase-specific, including photosynthesis, cell division and motility related
processes. By contrasting cyclic expression between C. reinhardtii and Arabidopsis thaliana
putative orthologs, we found significant but weak conservation in cyclic gene expression
patterns. On the other hand, within C. reinhardtii cyclic expression was preferentially maintained
between duplicates and the evolution of phase between paralogs is limited to relatively minor
time shifts. Finally, to better understand the cis regulatory basis of diel expression, putative cisregulatory elements were identified that could predict the expression phase of a subset of the
cyclic transcriptome. Our findings demonstrate both the prevalence of cycling genes as well as
the complex regulatory circuitry required to control cyclic expression in a green algal model,
highlighting the need to consider diel expression in studying algal molecular networks and in
future biotechnological applications.

18

INTRODUCTION
Diel (24 hour, day/night periods) cycles dictate physiological changes at different times
of day in many organisms. The timing of these physiological oscillations is regulated by a
combination of environmental, metabolic and circadian signaling processes (Farre 2012;
Kinmonth-Schultz et al. 2013; Song et al. 2013; Fonken and Nelson 2014). For example,
circadian clock mutants lead to phase changes under entrained diel conditions (i.e. light-dark
cycles) and changes in photoperiod sensitivity (Yanovsky and Kay 2002; McNabb and Truman
2008). Oscillations can be a direct adaptation to environmental cycles, for example restricting
photosynthesis and protection against UV radiation to periods of light. Diel cycles also influence
biotic responses such as defense mechanisms (Arimura et al. 2008; Goodspeed et al. 2012;
Baldwin and Meldau 2013) and mutualistic interactions (Frund et al. 2011; Lehmann et al.
2011). Mechanistically, many of these cycling responses are regulated at the transcriptional
level. For example, in the green alga Chlamydomonas reinhardtii, oscillations in starch levels are
partially regulated by the cyclic expression of ADP-Glucose pyrophosphorylase (Ral et al. 2006).
However, some circadian regulated processes are controlled at the post-transcriptional level
(Kojima et al. 2011) and/or by the interaction between transcriptional and post-translational
regulation (Kinmonth-Schultz et al. 2013; Song et al. 2013). Early transcriptome analyses of
three model organisms, Arabidopsis thaliana, Drosophila melanogaster, and Mus musculus,
indicated that between one and ten percent of genes exhibit circadian oscillation with periods of
~24 hr (Doherty and Kay 2010). Moreover, in photosynthetic organisms, 30-90% of genes cycle
under diel conditions (Michael et al. 2008; Monnier et al. 2010; Shi et al. 2010; Filichkin et al.
2011). In land plants, about a third of the genes that cycle in light/dark are also circadian
regulated (Michael et al. 2008; Filichkin et al. 2011). Several cis-regulatory elements (CREs)

19

necessary for circadian regulated gene expression have been identified (Michael and McClung
2002; Harmer and Kay 2005; Michael et al. 2008), although it remains an open question how
well the identified CREs explain global cyclic expression patterns.
The green alga C. reinhardtii has been used extensively to study physiological processes
under the control of circadian and/or diel cycle (Mittag et al. 2005; Matsuo and Ishiura 2010). C.
reinhardtii’s size, short life-cycle, and extensive genetic tool set make it an ideal model organism
(Harris 2001) particularly for studies such as experimental evolution from single to multicellularity (Ratcliff et al. 2013) and the genetic engineering of triacylglycerol accumulation in
algae (Grossman et al. 2007; Hu et al. 2008; Siaut et al. 2011). C. reinhardtii has also been used
to study rhythmic responses to light (Bruce 1970), ammonium (Byrne et al. 1992) and nitrogen
availability (Pajuelo et al. 1995). However, studies of cyclic expression in C. reinhardtii have
been limited to single (Mittag et al. 2005; Matsuo and Ishiura 2010) or relatively small sets of
genes (Kucho et al. 2005). Despite the large evolutionary distance, there are some conserved
elements between both the circadian (Corellou et al. 2009; Matsuo and Ishiura 2010) and the
photoperiodic (Romero and Valverde 2009) oscillators of flowering plants and green algae,
raising the question whether and to what extent cyclic expression is conserved. Therefore, a
genome wide analysis of cyclic expression in C. reinhardtii can provide insight not only into
cyclic physiological behavior in green algae, but also how this behavior has evolved in divergent
lineages of the Plantae. Such an analysis will also be relevant to economically important
processes in algae such as oil production.
In this study, we examined gene expression patterns under diel conditions in C.
reinhardtii. We characterized the prevalence of cycling gene expression in the C. reinhardtii
genome and observed that genes involved in distinct biological processes are consistently

20

expressed at certain times during the day/night cycle. We also investigated the conservation of
cyclic expression patterns between orthologs in C. reinhardtii and Arabidopsis thaliana, which
diverged ~650-800 million years ago (Sanderson et al. 2004) and the evolution of cycling
paralogous genes. Finally, to understand the cis-regulatory basis of diel expression, we identified
putative CREs (pCREs) associated with cyclic expression at different phases and investigated
how these pCREs can be used to predict cycling gene expression.

21

RESULTS AND DISCUSSION

Cycling gene expression is extensive in the C. reinhardtii genome
To characterize cyclic expression in C. reinhardtii, the expression profiles of 17,114
annotated C. reinhardtii genes were defined from samples taken at three hour intervals over two
24-hour time courses (see Methods). A gene was defined as cyclically expressed if it exhibited
statistically significant, non-random variation at a regular period as identified by either COSPOT
or DFT (see Methods). The union of predictions for both methods covered 8072 cyclically
expressed genes (47.2% of the C. reinhardtii genome), which we hereafter refer to as “cycling
genes”. Both approaches generated cyclic expression models that correlated with the original
expression data, with an average Pearson correlation coefficient of 0.987 for COSPOT and 0.880
for DFT. The correlation for COSPOT models is higher compared to that of DFT because
COSPOT models are fit directly to the overall pattern of the data while the DFT models are
based only on variations which occur at a period of 24-hours. Taken together, cyclic variation in
gene expression represented the predominant form of non-linear variation in RNA content at
both the genome wide and individual gene level.
Cyclic variation can be described using three parameters: period, amplitude, and phase
(Figure 2.1A). Using the fitted models, we inferred the period, amplitude and phase of all
cycling genes in the C. reinhardtii genome. The distribution of period for our set of cycling
genes was centered around 24 hours (+/- 1.10 hours, 95% confidence interval) (Figure 2.1B,
Supplemental Figure 2.1A). The amplitude of cyclic expression was highly correlated with
mean expression level (r2 > 0.7) and, on average, was only half the size of the mean, indicating
that most cycling genes are expressed at some constitutive level even during the trough of the
cycle (Figure 2.1C, Supplemental Figure 2.1B). The phase distribution of cycling genes was
22

Figure 2.1 Period, amplitude, and phase of cyclic expression. (A) Three properties of cyclic
variation: period, amplitude, and phase. (B) The distribution of period of cycling genes identified
in C. reinhardtii. (C) The relationship between amplitude and mean expression level in FPKM
(Fragments per Kilobase of transcript per Million mapped reads). (D) The distribution of the
phase of cycling genes.

23

bimodal with one peak at around ZT 0 (20.6% of cycling genes) and a second around ZT 12
(16.4% of cycling genes), corresponding to the night-to-day and the day-to-night transitions,
respectively (Figure 2.1D, Supplemental Figure 2.1C). Our finding concurs with the phase
distribution reported for A. thaliana and other plant species under diel conditions (Michael et al.
2008; Filichkin et al. 2011) as well as a subset of circadian genes in C. reinhardtii (Kucho et al.
2005).
Phases of cycling gene expression are associated with a succession of biological functions
Earlier studies have shown that multiple processes in C. reinhardtii have specific
rhythms, including the expression of key photosystem components (Hwang and Herrin 1994;
Jacobshagen and Johnson 1994) and the timing of gametogenesis (Jones 1970). Thus we first
asked which processes tend to be rhythmic by identifying GO terms with an over-represented
number of cycling genes. We found that cycling genes were enriched in 44 GO terms, including
those related to the chloroplast, photosynthesis, and ribosomal subunits (Supplemental Table
2.1). Among these terms, the most striking pattern was that 207 of 252 flagella related genes
showed cyclic expression. In particular, 80% of cyclically expressed flagella genes (167 of 207)
had peak expression at ZT 21, suggesting that biological functions can be phase specific. To
further explore the association between phase and function, cycling genes were assigned to eight
“phase clusters” (ZT 0, 3, 6, 9, 12, 15, 18, and 21; Figure 2.2A) and enrichment of GO
categories within each cluster was determined.
We found that 237 GO terms had over-represented numbers of genes in ≥1 phase cluster
(Figure 2.2B). Enrichment values for each term in each phase group can be found in
Supplemental File 2.1. The greatest number of enriched terms was found in the ZT 21 cluster,
just before the night-day transition, (40/237, 16.7%) and the ZT 9 cluster, just before the day-

24

Figure 2.2 Phase of gene expression and cyclically expressed GO terms. (A) The normalized
(relative) expression of each cycling gene in C. reinhardtii (each row) across the 48-hour period

25

Figure 2.2 (cont’d)
(columns). Genes were assigned to phase clusters based on the predicted time of peak
expression. Genes in each phase cluster were ordered using hierarchical clustering. The white
and black bars below indicate samples from the light and the dark periods, respectively. (B) The
test statistics of GO term (rows) enrichment in each phase (columns) in C. reinhardtii. The log(p-value) of the Fisher exact test is plotted. GO terms are ordered along the y-axis according
to the most enriched phases and hierarchically clustered within each phase. (C) The test statistics
of GO term enrichment in each phase in A. thaliana. Methods for assigning GO terms to phase,
clustering, and the color legend are the same as in (B).

26

night transition (61/237, 25.7%). We also observed that over-represented GO terms tended to be
phase-specific: of all 237 terms, only 19 were enriched in >1 phase and 12 of those were
enriched only in two adjacent phases (Figure 2.2B, Supplemental Figure 2.2A). In contrast, the
majority of under-represented categories (51%) spanned ≥ 4 phases (Supplemental Figure
2.2B). Thus, genes involved in the same process not only tended to be enriched in a particular
phase of expression, but were also depleted in other phases. This phase-specificity of functional
categories was consistent with previous studies of light-response, metabolism, cell division, and
flagellum biogenesis in C. reinhardtii demonstrating cyclic behavior at a specific time of the day
(Jones 1970; Cavalier-Smith 1974; Teramoto et al. 2002). For example, DNA replication and
mitotic events in C. reinhardtii are restricted to the early hours of the dark period (Jones 1970):
not only is the transition into darkness required for normal cell division (Voight and Munzer
1987), but DNA replication and cell separation occur between 2-5 hours after the light-dark
transition (Fang et al. 2006). This specific timing of DNA replication after the light to dark
transition matches the phase of expression for cycling genes related to this process.
Alternatively, the gradual increase in expression of replication associated genes towards a peak
early in the dark period may track with increases in cell size, as it has been shown that the
concentration of cell cycle regulatory proteins HA-MAT3, DP1, and E2F1 remain constant in
spite of the increase in cell volume during G1 (Olson et al. 2010). We should note that many of
the phase-specific functional categories uncovered here, such as amino acid biosynthesis,
phosphorelay activity, and mRNA splicing were not previously known to show time-specific
cycling behavior in C. reinhardtii. While correlation alone is insufficient to prove causation, the
coordination between cyclic expression and function is highly suggestive that timing of
transcription can regulate the timing of higher order biological processes.

27

Based on the apparent association between phase and function in this as well as in prior
studies, GO terms were classified into broad “functional groups”: (1) ribosome and translation,
(2) photosynthesis and light response, (3) mitochondria and metabolism, (4) cell cycle and
mitosis and (5) microtubules and flagella (Figure 2.3, Supplemental Table 2.2). We found that
group 1, 2, and 3 were over-represented in the middle of the day (ZT 3 and 6), group 4 in the
early and mid-night (ZT 12 and 15), and group 5 at the end of the dark-period (ZT 21) (Figure
2.3A). Consistent with the pattern of phase-specific enrichment of genes in different functional
groups, the normalized expression profiles of cycling genes in each functional group clearly
demonstrated phase specificity (Figure 2.3B-F). The diel expression data also highlighted the
possibility of distinguishing different components of a biological process. For example, group 5
genes are involved in forming microtubules and subsequently flagella. Within this group, genes
associated with the microtubule cytoskeleton peaked earlier in the dark period while those
associated with flagellum assembly peaked toward the end (Figure 2.3L), representing a clear
delineation between spindle body formation and flagellar regeneration as described previously
(Cavalier-Smith 1974). Taken together, our findings suggest that the timing of biological
processes (translation, cell-replication, and regeneration of the flagellum) may be determined by
transcriptional regulation.
C. reinhardtii and A. thaliana orthologs have limited conservation in cycling gene
expression patterns
To test if the functional coordination and phase specificity of cyclic expression observed
in C. reinhardtii can be found in related multicellular species, cycling genes were identified in A.
thaliana using the same methods and cutoff values applied to C. reinhardtii on an existing diel
expression data (Blasing et al. 2005). A total of 4945 genes in A. thaliana were identified as

28

Figure 2.3 Phase specific expression of broad functional categories. (A) Enrichment test
statistics in each functional group (row) and in each phase cluster (column) among C. reinhardtii
29

Figure 2.3 (cont’d)
cycling genes. The color indicates the averaged –log(p-value) of GO terms in a functional group
(Supplemental Table 2.2). (B-F) Normalized expression profiles of genes in each functional
group in C. reinhardtii. The black line indicates average expression values. The grey area
represents plus/minus one standard deviation. (B) Ribosomes/Translation (C)
Photosynthesis/Light-response (D) Mitochondria/Metabolism (E) Cell-cycle/Mitosis (F)
Microtubules/Flagella (G) Enrichment test statistics for functional groups in A. thaliana. The
functional group designation and color legends are the same as (A). Gray: not applicable. (H-K)
Normalized expression profiles of genes in each functional group in A. thaliana. (H)
Ribosomes/Translation (I) Photosynthesis/Light-response (J) Mitochondria/Metabolism (K) Cellcycle/Mitosis (L) Expression profiles of genes in the microtubule cytoskeleton (red), flagellum
assembly (blue) and cell projection organization (black) categories.

30

cycling (21.7% of the annotated genes), less than half of what was seen in C. reinhardtii. This
difference is in part due to a lower sampling density of the A. thaliana data (once every 4 hours),
though the overall time span covered was longer (3 days). It is also possible that the mixture of
different cell types in A. thaliana samples could mask some rhythmic expression patterns. We
also observed that 992 GO terms in A. thaliana were over-represented in ≥1 phases compared to
237 in C. reinhardtii, which is likely a function of significantly better annotation (Figure 2.2C).
Enrichment values for each term in each phase group can be found in Supplemental File 2.2.
In contrast to the strict phase-specificity in C. reinhardtii, A. thaliana group 2 GO terms
(photosynthesis and light response) were enriched amongst cycling genes in all six time points,
but were predominant at ZT 4. The other three groups (group 1, 3, and 4) were restricted to at
most two adjacent phases (Figure 2.3G). Compared to C. reinhardtii, there is a greater variance
in the phase of expression amongst the A. thaliana cycling genes within each group, potentially
due to the fact that the A. thaliana expression data was derived from samples of mixed tissues
and cell types. Nonetheless, the peak expression of photosynthetic, mitochondrial, and
ribosomal genes occurred at a similar time, as was observed in C. reinhardtii (Figure 2.3H-K).
These results suggest that cyclic expression is conserved between a subset of functionally related
genes, in both unicellular and multi-cellular plant systems.
Due to the concern that the phase-specificity differences between C. reinhardtii and A.
thaliana might be due to annotation quality difference, we next examined the degree to which
cycling gene expression was conserved between orthologous genes in these two species. Among
11,845 putative orthologs, 1,464 (12.4%) showed cyclic expression in both species (referred to
as “co-cycling” orthologs), which is significantly higher than the random expectation (ChiSquared Test, pv < 0.001). The conserved co-cycling genes encode components of the ribosome

31

(particularly the small subunit), plasma and thylakoid membrane components, or are involved in
stress response (Fisher Exact Tests, pv < 0.05). Nonetheless, we should emphasize that the
difference between the observed and expected proportion of co-cycling orthologs was only 2.4%.
Thus most cycling genes in C. reinhardtii are not cyclic in A. thaliana and vice versa. In
addition, while the amplitude of cyclic expression is significantly correlated among co-cycling
orthologs (r2 = 0.30, pv < 10-100), there are only weak relationships between their phases (r2 <
0.01, pv < 0.006). The A. thaliana and C. reinhardtii lineages diverged 650-800 million years
ago (Sanderson et al. 2004) and have extensive differences in life histories, distribution,
complexity, and physiology. Thus the conserved components of cyclic expression are likely core
processes strongly selected to be maintained, including photosynthetic, mitochondrial, and
ribosomal genes (Figure 2.3H-K). However, most orthologs between green algae and flowering
plants have divergent patterns of cyclic expression, and the extent of cyclic expression
divergence highlights the fact that cycling gene expression can be plastic.
Conservation of cyclic expression is more prevalent amongst older duplicate genes
To further assess how quickly cyclic expression divergence occurred, we asked how the
pattern of cycling gene expression evolved between duplicated genes in C. reinhardtii. Gene
trees were inferred based on similarity of known protein domains and we retained only the closet
pairs of paralogs (i.e. those separated by only a single ancestral node) for subsequent study (see
Methods). The frequency with which the pattern of gene expression (cycling or non-cycling) was
identical or divergent was compared to the timing of the inferred duplication event, estimated
using the synonymous substitution rate (Ks) (Figure 2.4A). The overall frequency of diverged
duplicates (one paralog cycling, the other non-cycling) increased with Ks, approaching an
asymptote of ~0.45 for Ks > 0.9. While the frequency of non-cycling duplicates decreased with

32

Figure 2.4 Conservation of cyclic expression and phase of cyclic expression. (A) The
frequency at which duplicate pairs of genes in C. reinhardtii maintain cycling expression,
maintain non-cycling expression, or diverge as a function of the synonymous substitution rate
(Ks). (B) Distribution of cycling retention, non-cycling retention and divergence between
duplicate pairs in random simulations. The black bars cover the inter-quartile range of each
distribution, and error bars represent the 95% confidence interval. Observed values are indicated
by asterisks. (C) The frequency at which the phase is retained in pairs of cycling duplicates as a

33

Figure 2.4 (cont’d)
function of Ks. (D) Enrichment values for phase retention (diagonal values) and phase change
(off diagonal values) between actual duplicates and duplicate pairs in random simulations.

34

Ks, the frequency of cycling duplicates was greater on average for Ks > 0.9 indicating a net gain
of cycling expression as duplicates age. We hypothesized that this gain of cycling expression
results from a bias in the rate at which duplicate genes diverge that favors the cycling state.
To test this hypothesis, we examined if the observed changes in the frequency of
retention can be explained without assuming different rates of divergence. Therefore, a null
model of duplicate gene divergence was created using a system of difference equations (see
Methods). We fit the transition probabilities using the difference in frequencies between Ks 0.6
and 0.9, and the predicted frequencies of identical and divergent duplicates closely matched our
observed results at all time points (root mean squared error = 0.03), showing the same pattern of
increases and decreases (Supplemental Figure 2.3). Hence, we have no evidence of a
differential rate in divergence between cycling and non-cycling duplicates, however the
predicted probability of transition from identical to divergent (0.42) is less than the probability of
transition from divergent to identical (0.53), suggesting that there is a preference for maintaining
duplicates in an identical state. This is consistent with our finding that the observed frequency of
paralogs with identical states tends to be significantly higher than expected under random
association (Z-test, pv < 10-17; Figure 2.4B). In contrast, the frequency of paralogs with
divergent state is significantly lower than expected (Figure 2.4B).
Next, we examined the frequency with which phase is identical amongst pairs of the
cycling duplicates. Overall, the number of co-cycling paralogs for which the phase of cyclic
expression was identical is more than twice the number randomly expected (Z-test, pv < 10-85)
with 33.7% of co-cycling duplicates sharing the same phase. The identical phase state was more
common amongst cycling duplicates with lower Ks and there was a sharp decrease in the
frequency of duplicates with identical phases going from a Ks of 0.9 to 1.2 (Figure 2.4C). Next

35

we explored if there was a bias in the magnitude of phase change between co-cycling duplicates
(Figure 2.4D). We found that small phase divergences of +/- 3 hr (covering 28.3% of all
duplicates) tended to be enriched relative to random expectation, in particular at ZT0/ZT3 and
ZT12/ZT15, although the identical phase state is still the most highly enriched scenario.
Additionally, there was an inverse, linear relationship between the magnitude of the difference in
phase between cycling duplicates and the enrichment of phase-shift events relative to random
expectation (all cycling duplicates, r2 = 0.91; duplicates with Ks > 0.9, r2 = 0.93), indicating that
large differences in phase between duplicates occur less frequently than expected by random
chance. Furthermore, we found 33 GO terms enriched (adjusted p-value < 0.05) amongst cycling
duplicates with the same phase, the majority of which (88%) were previously found to be
enriched in a specific phase of cyclic expression.
Cycling genes are enriched for specific putative Cis-regulatory elements
The coordinated expression of functionally related genes suggests the existence of one or
more regulatory mechanisms that drive phase specific expression. While mRNA levels may be
affected at multiple levels of regulation, we chose to focus on transcriptional regulation driven
by cis-regulatory sequences as circadian rhythm related cis-elements have previously been
identified in plant and animal models (Michael and McClung 2002; Ueda et al. 2005; Michael et
al. 2008). Using a motif finding pipeline (Zou et al. 2011), we found 687 putative cis-regulatory
elements (pCREs) in the 1kb regions upstream of the transcriptional start sites of cycling genes
for each of the eight C. reinhardtii phase clusters (Fisher Exact Test, adjusted pv < 0.05). The top
enriched motifs for each phase can be found in Figure 2.5, and the entire list of enriched motifs
can be found in Supplemental File 2.3. Each phase had 60-84 associated pCREs, except for ZT
15 with 169; however, more than 20% of pCREs (141/687) were enriched in >1 phase and 43.8%

36

of ZT 15 pCREs (74/169) were enriched in ≥1 other phases (mostly ZT 12; Figure 2.6A).
Therefore, each pCRE was assigned to the phase cluster in which it was most significantly
enriched.
To further assess whether the pCREs are meaningful, they were used to establish
classifiers to predict cyclic expression in different phases. First, the pCREs assigned to each
phase were used to predict which genes are cyclic in a naĂŻve manner. That is, for pCREs
enriched in a particular phase, we simply predicted that all genes with ≥1 pCREs mapped to their
promoters would cycle at that phase. The performance of these predictions was assessed using
the area under the receiver operating curve (AUC-ROC), a metric which quantifies the ability of
a method to predict positive examples which, in our case, is phase specific expression. Perfect
predictors have an AUC-ROC of 1 while random guessing has a value of 0.5; our naĂŻve
classification of phase had AUC-ROCs that ranged from 0.58 (ZT 9) to 0.62 (ZT 12) indicating
that this simple classification procedure performed marginally better than randomly assigning
phase (Figure 2.6B). The same conclusion can be reached based on the F-measure, another
prediction performance metric (Figure 2.6C). Next, to further improve the prediction of the
phase of cyclic expression, we used the support vector machine (SVM) algorithm to classify
cycling genes according to the presence or absence of all pCREs (see Methods). The SVM
classifier shows improved performance compared to naĂŻve classification (Figure 2.6B-C,
Supplemental Table 2.3) but AUC-ROC values are still relatively low, ranging from 0.58 (ZT
9) to 0.65 (ZT18) (Supplemental Figure 2.4). We also identified two pCRE association rules
enriched in specific phases of cyclic expression using CBA (Liu et al. 1998); however, adding
these rules to the SVM prediction models did not significantly improve the overall predictive
power of our pCREs as the AUC-ROC increased by at most 0.01.

37

Figure 2.5 Top three pCREs enriched in each phase cluster of cyclic genes. Sequence logos
representing the top three putative cis regulatory elements (pCREs) enriched in each phase
cluster of cycling genes in C. reinhardtii.

38

Figure 2.6 Enrichment and performance of phase-specific pCREs. (A) The enrichment test
statistics of 687 pCREs (rows) in genes of each phase cluster and non-cyclic (NC) genes
(columns). (B) The area under the curve of the receiver operating characteristic (AUC-ROC) for
phase expression prediction with naĂŻve (green) and Support Vector Machine (SVM, white)
classifiers. (C) The F-measures for phase expression prediction based on random guess (black),
naĂŻve (green) and SVM (white) classifiers.

39

Given that the C. reinhardtii pCREs are computationally derived, we next asked how
well a known, experimentally verified, phase-specific cis-regulatory element may predict cyclic
expression. For this purpose we examined the Evening Element that is necessary and sufficient
to drive circadian expression in A. thaliana (Michael and McClung 2002; Harmer and Kay 2005;
Michael et al. 2008). Using motifs related to the Evening Element identified in Michael et al.
(2008), we generated a cycling gene classifier to predict the phase of the 4,945 A. thaliana
cycling genes introduced in the earlier phase-specificity comparison section. The optimal AUCROC of the Evening Element classifier was 0.57 and 0.56 at ZT 0 and 12 hours, respectively
(compared to 0.58-0.65 in C. reinhardtii pCRE predictions). Therefore, although the Evening
Element is known to function as a circadian regulator, similar to C. reinhardtii pCRES, it has
only limited predictive power on a genome wide scale. To obtain accurate predictions the
presence or absence of pCREs needs to be supplemented with additional information regarding
the regulation of cycling expression.
Phase of cyclic expression can be predicted for groups of genes with common expression
patterns or common function
The weak predictive power of pCREs likely results from an underlying complexity in the
regulation of the phase of cyclic expression, either in the form of additional control mechanisms
or the existence of more discrete regulatory groups. Timing of cyclic expression may be
modified by interactions amongst regulatory motifs or post-transcriptional mechanisms. It is also
possible that our phase clusters might consist of multiple regulatory subgroups. To address the
latter possibility, we further classified genes in each phase group into sub-clusters containing
genes with highly similar expression profiles (phase-expression clusters). Using SVM, 28 of 190
phase-expression clusters covering 584 genes (7.23% of cycling genes) could be classified with

40

an AUC-ROC > 0.7 (these clusters are described in Supplemental File 2.4), which is better than
any individual phase alone. The best predicted phase-expression clusters do not necessarily have
stronger cyclic signals (Figure 2.7A) compared to the worst predicted (Figure 2.7B).
Additionally, we eliminated size (r2 = 0.15) and the correlation of expression profiles within each
phase-expression cluster (r2 < 0.01) as possible variables explaining the observed variance in
AUC-ROC (Supplemental Figure 2.5). These results suggest that phase specific regulation does
occur at the cis-regulatory level for particular groups of cycling genes and that presence or
absence of pCREs alone is sufficient to accurately predict the pattern of phase specific
expression for these clusters. Those pCREs which were informative (i.e. had the non-zero
weights) when predicting the 28 best phase-expression clusters are listed in Supplemental File
2.5.
In addition to using highly similar expression patterns as a way of subdividing phase
clusters, we looked for evidence of phase specific regulation amongst groups of genes in the
same phase cluster that had related annotated function (phase-function clusters). Among 71
phase-function clusters, genes belonging to 12 of these clusters could be classified with an AUCROC > 0.7. These clusters covered 12.2% (175/1434) of genes present in all phase-function
clusters, a higher percentage than the phase-expression clusters, although they constitute a
smaller portion of all cycling genes due to limited GO annotation in C. reinhardtii (these cluster
are described in Supplemental File 2.6). Genes in most of these functional groups displayed a
clear cyclic signal (Figure 2.7C), except for the groups related to the nucleolus and cell wall,
which were predominantly non-cyclic genes but had a statistically significant subset of phasespecific genes. Amongst the best classified sub-clusters contained genes relating to the large
cytosolic ribosomal subunits (AUC-ROC = 0.73), cilium (0.72), small cytosolic ribosomal

41

Figure 2.7 Expression of best predicted co-expression cluster and GO terms. (A) Averaged,
normalized expression profile of genes in the top 28 co-expression clusters whose phase of
expression can be predicted with AUC-ROC > 0.7. (B) Averaged, normalized expression profiles
of genes in the bottom 28 co-expression clusters whose phase of expression can be predicted
with AUC-ROC < 0.56. (C) Averaged, normalized expression profiles of genes in the 12 GO
terms whose phase of expression can be predicted with AUC-ROC > 0.7. Both cycling and noncycling genes annotated in each GO term are included.

42

subunit (0.72), translation (0.71), and the chloroplast (0.68). This supports our earlier observation
that the cyclic patterning of large scale processes such as photosynthesis, translation, and
motility may be regulated at the transcriptional level. The pCREs which had non-zero weights
when predicting the 12 best phase-function clusters are listed in Supplemental File 2.7.

43

CONCLUSIONS
We have determined that cyclic expression is prevalent in the C. reinhardtii genome, and
nearly half of all annotated genes cycle under diel conditions. There is a strong link between
rhythmic patterns at the molecular and physiological levels. Diel cycling expression is influenced
both by environmental factors, such as the availability of light, and endogenous factors,
including metabolism and the circadian clock (Farre 2012; Kinmoth-Schultz et al. 2013; Song et
al. 2013; Fonken and Nelson 2014). While the importance of photoperiod can be inferred for
light-dependent (i.e. photo-synthesis) and light-sensitive (i.e. DNA replication) processes, for
most cycling related functions it remains an open question as to what extent each factor
influences cycling expression. This is particularly true of functions which were not previously
known to exhibit cycling expression in green algae, for example, the regulation of RNA
processing and amino-acid synthesis.
In addition to the relationship between cyclic expression and gene function, we found that
cyclic expression was significantly conserved between paralogous genes. The proportion of
divergent duplicates reaches an asymptote at Ks > 0.9, which is similar to what was previously
observed for stress responsive duplicate genes (Zou et al. 2009). However, while there appears to
be a clear preference for the partitioning of ancestral expression states in stress responsive genes
(Zou et al. 2009; Dong and Adams 2011), we found that duplicates genes tend to share the same
expression state with respect to cycling and that cycling duplicates preferentially retain the same
or similar phase of expression. We hypothesize this pattern of cyclicity/phase conservation
among duplicates points to a fundamentally distinct regulatory logic from that of stress response.
In stress response, a duplicate which has lost response to one condition may still be responsive to
other conditions and thus retained. However, either loss or gain of cyclicity in a duplicate gene

44

would mean it is no longer temporally in sync with other genes in the processes which it was
originally involved in. For example, if a replication initiation factor duplicate was not in sync
with the expression of other components of the replication machinery, the duplicated factor
would not be functional and eventually eliminated from the genome. This argument may also
apply to the conservation of phase among duplicate cycling genes. Indeed, we found that most
GO terms enriched amongst co-cycling duplicates with the same phase were highly phase
specific, including those associated with DNA replication and flagellar components.
Based on prior studies of stress response genes (Zou et al. 2009), we expected that the
conservation of cycling expression state, particularly the phase of expression, would be
correlated with the presence of shared cis-regulatory elements. However, contrary to this
expectation, the set of putative cis-regulatory elements enriched in cycling genes does not
accurately distinguish phase expression. While our results suggest that cis-regulation plays a
significant role in controlling cyclic expression in C. reinhardtii, the presence or absence of
promoter elements alone was insufficient to fully explain the observed patterns of cyclic
variation across the entire C. reinhardtii genome. This suggests that additional regulatory
components are involved in controlling cyclic expression. In other organisms the combinatorial
interactions amongst regulatory factors play an important role in controlling the phase of cyclic
gene expression (Harmer and Kay 2005; Ueda et al. 2005), but in C. reinhardtii there is evidence
that response to changing light levels is mediated by multiple copies of the same or similar
promoter elements (von Gromoff et al. 2006). While we did not see significant improvement
when rules considering combinatorial relationships between pCREs were included in our model,
this may be due to the fact that we were able to explore only a subset of all possible
combinatorial interactions in our pCRE set. Additionally, post-transcriptional regulation has been

45

implicated in regulating circadian processes in Neurospora crassa, A. thaliana, and D.
melanogaster (Kojima et al. 2011). In C. reinhardtii, the over or under expression of the RNAbinding protein CHLAMY1 is known to result in the disruption or loss of circadian rhythms
(Iliev et al. 2006). Further studies incorporating post-transcriptional regulatory features will be
necessary to improve the prediction of phase specific cyclic expression
The inability of pCREs to classify phase specific cycling expression on a genome wide
scale does not contradict prior observations that certain cis-elements are necessary for cycling
expression (Michael and McClung 2002; Harmer and Kay 2005; Michael et al. 2008). Rather it
suggests that cis-elements alone are insufficient to explain the variation in cycling expression on
a genome wide scale and that additional regulatory components remain to be discovered. Posttranscriptional regulatory mechanisms and chromatin state are two promising avenues of
investigation which, in conjunction with the cis elements we have identified, could be used to
better predict the state of cycling expression. Although there remains substantial room for further
improvement, our findings contribute to a better understanding of both the function and
evolutionary origins of cyclic expression in a green alga, laying the foundation for further
molecular dissection of the relationships between the rhythmic gene expression and
physiological functions for potential biotechnological applications.

46

MATERIAL AND METHODS

Growth of Chlamydomonas reinhardtii Cultures
C. reinhardtii dw15.1 was grown in TAP (Tris-Acetate-Phosphate) media in flasks
without aeration, shaken at 100 rpm, at 22 °C. While the acetate present in this media provides
an alternative source of carbon, allowing for C. reinhardtii to grow in the dark, prior studies have
shown that that the cell cycle (Voight and Munzner 1987; Davies and Grossman 1994) and other
metabolic cycles (Ral et al. 2006) are still synchronized in C. reinhardtii grown in acetatecontaining media under light/dark cycles. Additionally, the amplitude and phase of cell cycle
gene expression in our study and in previous studies where cultures were grown under
autotrophic conditions (Bisova et al. 2005) are similar (Supplemental Figure 2.6). An initial
200 mL culture was grown to a density of 25 million cells mL-1 in constant light (50 mol s-1 m2

) and used to set up 50 mL cultures of 0.5 million cells mL-1 that were transferred to 12 hr light

(50 mol s-1 m-2) and 12 hr dark conditions for 48 hours prior to sampling. Two biological
replicates were collected every 3 hours between ZT (Zeitgeber Time, hours since last dawn) 0
and ZT 21. Each sample originated from an independent 50 mL culture. Samples collected
during the light to dark or dark to light transition were taken just prior to change of conditions.
For collection, 2 mL of the culture was placed in a 2 mL tube and centrifuged at max speed in at
4°C for 10 min. Amber tubes were used for samples collected during the dark period and the
supernatant was removed under weak green light. The pellets were snap frozen in liquid
nitrogen. The frozen samples were ground using the Qiagen tissue lyser for RNA extraction.

47

RNA-sequencing
RNA was extracted using the Omega eZNA Plant RNA kit. The RNA was eluted in 50
L DEPC-H2O and the concentration was measured using a Nanodrop (Thermo-Fisher). A
portion of the RNA was diluted to 1 ng uL-1 to check the RNA Integrity Number (RIN) with a
Bioanalyzer (Agilent). All samples had a RIN equal to or greater than 7. Library preparation and
sequencing was performed at the MSU-Research Technology Support Facility using the Illumina
Tru-Seq Stranded kit with an Illumina HiSeq 2500. Eight samples were sequenced in each lane
using a custom bar-coding, but the two biological replicates from the same time point were run
in separate lanes. The average number of RNA-Seq reads per sample was 1.81e7 and they ranged
between 7.07e6 and 2.58e7. The reads from each of 16 samples (8 time points, 2 samples each
time point) were mapped to the C. reinhardtii genome (version 4.3 from Phytozome) using
Tophat (Trapnell et al. 2009) with default parameters except for intron length (min 13, max
8712) and max-multi-hits (1). Gene models on non-chromosomal fragments were not considered.
FPKM (Fragments Per Kilobase of transcript per Million mapped reads) per gene was calculated
using Cufflinks (Trapnell et al. 2010) with parameter –I 8712. A high percentage of reads
mapped to the genome: the least mapped sample had 82% of reads mapped and the average of all
samples was 85%. Upper quartile normalization was applied to all samples to correct for
technical variation as recommend in (Bullard et al. 2010). The two biological replicates were
appended and used as two consecutive days for downstream analysis. Raw read data is available
through the NCBI SRA, BioProject accession [PRJNA264777].
Identification of Cycling Genes
Two programs were used to identify cyclic patterns of expression in FPKM data:
COSPOT (which is described in (Panda et al. 2002)) and an application of the discrete Fourier

48

transform (DFT). The DFT has previously been applied to the analysis of cyclic expression using
RNA-Seq data (Rodriguez et al. 2013), but our method is based primarily on PRIISM (Rosa et
al. 2012). We chose to use both COSPOT and the DFT in conjunction because we found that the
combination of methods had superior coverage of known cycling genes without a substantial
increase in the expected false positive rate (see Supplemental Materials and Methods)
In our application, we take the discrete Fourier transform of each gene expression vector
in the C. reinhardtii FPKM data set, converting a set of ‘N’ FPKM values (x) in terms
expression vs. time to new values (y) in terms of expression vs. frequency such that:
N 1

yk   xn * e

i 2

kn
N

(1)

n 0

Where xn is the FKPM value at the nth time point and yk is the kth frequency component with
period T/k where T is the time period spanned by the expression vector. The set of frequency
components represents the power spectra of the associated expression vector, that is, the
contribution of each periodic cycle to the overall data. In calculating the power spectra of the
expression data, we employed a non-windowed application of Welch’s Method (too few data
points were present tolerate the loss of information involved in windowing) to average the power
spectra over subsets of the expression vector with T = 24 hr. This was done to reduce bias in the
calculation of the power spectra that might be induced by a particular subset of the expression
data at the cost of reducing the overall resolution of the power spectra (though this loss of
resolution was primarily at the extreme ends of the spectra and should not affect our results).
Furthermore, the coefficients of each power spectra were normalized prior to averaging using the
following equation:
yk* 

yk  ymin
ymax  ymin

49

(2)

Where ymin is the smallest coefficient of the power spectra and ymax is the largest. As such, the
normalized values, yk* , are on the interval [0,1], further reducing the affect that a single subset
can have on the average power spectra. The “cyclic score” of each gene is defined as the
normalized value of the 24 hour frequency component. This score is equated to a p-value by
randomizing the order of values in each expression vector and scoring the vectors in this random
population. For this study, we tested cyclic score thresholds equal to the 5th, 2nd, and 1st
percentiles of the score distribution of the randomized data (equal to 0.745, 0.808, and 0.841
respectively) and chose the 2nd percentile as our cutoff for calling cycling genes (equivalent to a
p-value of 0.02). In comparison, the 5th percentile of cyclic score for the set of predicted cycling
genes in C. reinhardtii was 1. Additional information about how these thresholds were
determined as well as a comparison to COPSPOT can be found in the Supplemental Materials
and Methods.
Clustering Cycling Genes According to Phase
Cycling genes in C. reinhardtii were first divided by their phase of expression, that is, the
Zeitgeber Time (ZT) at which peak expression occurred in the FPKM data set. Within each
phase cluster, genes were ordered using hierarchical clustering implemented in R for display
purposes. Phase clusters were further broken down using two-rounds of k-means clustering,
implemented using custom Python scripts. K-means clustering involves initially selecting “k”
random centers in parameter space and assigning genes to clusters based on their distance to the
nearest center. The mean of each cluster is then used to define new centers which in turn are used
to redefine clusters; this process is repeated until the clusters converge or the amount change per
iteration falls below a specified threshold. The final clusters used for pCRE identification
contained 10 to 90 genes. Enrichment of Gene Ontology (GO) terms and pCREs in phase groups

50

was done using the Fisher Exact Test and the resulting p-values were corrected for multiple
hypothesis testing using the Benjamini-Hochberg method (Bengamini and Hochberg 1995).
Conservation of cyclic expression and phase of expression amongst duplicate genes
Gene trees in C. reinhardtii were defined using the pipeline described in Zou et al. (2009)
using a set of protein domains defined using PFAM (Punta et al. 2012). These domains were
extracted from protein sequences and aligned using MAFFT (Katoh et al. 2002), and a
phylogeny was inferred using RAxML (Stamatakis 2006) with parameters -f d -m
PROTGAMMAJTT. Large domain families were divided by building neighbor joining trees with
PHYLIP (Felesenstein 2005) and cutting at a distance to root ≥ 0.05 to create sub-clusters
between 4 and 300 genes in size. Domains were mapped back to C. reinhardtii genes to infer
gene trees. The gene trees, including the divided trees for large domain families, were reconciled
with an existing species tree (Moreau et al. 2012) using NOTUNG (Chen et al. 2000). An
archive of these gene trees in Nexus (.nex) format has been included as Supplemental File 2.8.
Branches containing A. thaliana and C. reinhardtii genes were extracted from the overall tree.
The significance of the retention rate of cyclic expression and the phase of cyclic expression was
determined by randomly pairing genes in the set of duplicates 100,000 times and comparing
retention among actual duplicates to the random population.
Modeling cycling state divergence of duplicate genes
The divergence of duplicate genes was modeled using a system of three difference
equations with a common rate ‘d’ for the divergence of both cycling and non-cycling duplicates
and a common rate ‘s’ for the reversion of diverged duplicates back to an identical state.
Duplicate gene pairs were binned according to Ks (width = 0.3), and we assume that the initial
frequency of duplicates was the same within each bin (If the initial conditions were significantly

51

different, we would expect to see deviation from the observed frequencies in the model
predictions, which was not the case). We then solved for values of ‘d’ and ‘s’ using the observed
change between consecutive bins, arriving at a solution with the same qualitative behavior as the
observed data. A detailed description of the model can be found in the Supporting Information.
Identification of putative cis-regulatory elements and phase prediction
Identification of pCREs in the promoter regions of C. reinhardtii genes followed the
pipeline described in Zou et al. (2011). Cycling genes were clustered according to phase and
expression profile as previously described. For each cycling gene the promoter region, defined as
the first 1kb upstream of the transcription start site less any bases which overlap with another
gene, was isolated. Six motif finders, AlignAce, MDscan MEME, Motif Sampler, Weeder, and
YMF, were used to identify motifs enriched in the promoter region of each phase cluster
compared to the promoters of all cycling genes. The resulting motifs were merged using
UPGMA to reduce the number of motifs and remove redundant motifs. Merged motifs were
mapped back to the C. reinhardtii genome using a threshold p-value of 1e-05.
The presence or absence of pCREs was used to predict the phase of expression of cycling
genes using a Support Vector Machine (SVM) implemented in Weka (Hall et al. 2009). Given a
test-set of positive and negative examples defined using n-variables (in this case, presence or
absence of pCREs), SVM seeks to define a linear classifier (i.e. a hyperplane in variable space),
which best divides positive and negative examples. This classifier is then used to assign
subsequent data points to either the positive or negative set. A grid search of two parameters, the
minimum distance between positive and negative groups (C) and the ratio of negative to positive
examples in the training set (R) were used to optimize separation and pick the best classifier. The
tested range of each parameter was as follows: C = (0.01, 0.1, 0.5, 1, 1.5, 2.0) and R = (0.25, 0.5,

52

1, 1.5, 2, 2.5, 3, 3.5, 4). Results were validated using 10-fold cross validation, which involved
dividing positive and negative examples for each phase into training test sets using stratified
random sampling. Each of the 10 test sets was classified by an independent SVM run and the
average of the 10 runs was used to score the performance of the parameter set.
Identifying groups of genes with common expression or common function
Cyclic genes with common expression were defined using k-means clustering as
described above. Cyclic genes with common function were defined as those which shared the
same GO annotation. For the purpose of predicting cyclic expression, we used only those GO
annotations over-enriched in at least one phase of cyclic expression and where at least 8
annotated genes were over-enriched in the same phase.

53

ACKNOWLEDGEMENTS
We thank Alexander Seddon and John Lloyd for help in identifying pCREs. We also thank all
the members of the algae group at Michigan State University for their attention and critique. This
work was supported by a Michigan State University Strategic Partnership Grant to C.B, E. F.,
and S.-H. S. and a National Science Foundation grant (MCB-1119778) to S.-H. S.

54

APPENDIX

55

Supplemental Materials and Methods
Determining threshold scores for COSPOT and DFT
Mittag et al. (2005) lists 18 proteins in C. reinhardtii which have previously show to
exhibit circadian changes in the rate of transcription or concentration of mRNA. Amino acids
sequences of these proteins were identified through KEGG and mapped to the C. reinhardtii
genome using the TBLASTN tool available through Phytozome. We found fifteen proteins
which mapped unambiguously to the C. reinhardtii genome and had matching annotation, only
one of which was not present in the mRNA seq data set (Supplemental Table 2.4).
To define a cutoff threshold for each of our methods, COSPOT and the DFT, each
program was evaluated against our gold-standard set three different p-values thresholds (or the
equivalent cyclic-score): 0.05, 0.02, and 0.01. For each p-value threshold, the coverage of both
the gold-standard set and the whole C. reinhardtii genome is reported in Supplemental Table
2.5. At each p-value threshold, the union of predictions from methods was used to define
cyclically expressed genes in C. reinhardtii. As such, the p-value of the new two-dimensional
threshold is defined by the joint distribution of COSPOT and DFT scores. Calculating this value
is complicated by the fact that these scores are highly correlated (r2 > 0.7), but the joint
probability can be estimated using a randomized population of expression vectors
(Supplemental Table 2.6). For every test p-value threshold, the increase in joint probability
(compared to individual significant thresholds) was relatively moderate whereas coverage of the
gold standard increased by as much 20% over a single method. We chose to use the combination
of COSPOT&DFT as our predictive method with a test p-value threshold of 0.02, which
balances in the inflation of the joint probability with the coverage of the gold standard set.

56

The combined method is most effective at excluding non-cycling genes, rather than
defining cycling genes, which can be seen by looking at the correlation of both methods at
different scoring threshold (Supplemental Figure 2.7). While the overall correlation between
both methods is high, the correlation amongst highly scoring genes (exceeding the 0.02 threshold
for either method) is actually quite low (r2 < 0.2). Genes which score very highly with one
method may be at or just below the margin for the other, however, a gene which scores poorly in
one method generally scores poorly with the other. Therefore, we chose a more conservative
score threshold as a cautionary measure.
Derivation of the model of duplicate gene divergence
Divergence of expression state was modeled using the following system of difference equations:

Ct 1  Ct  (1  d ) 

Dt s
2

(3)

Nt 1  Nt  (1  d ) 

Dt s
2

(4)

Dt 1  Dt  (1  s)  Ct d  Nt d (5)

Where C, N, and D represent the frequencies of cycling, non-cycling, and divergent
duplicates at a given Ks (subscript “t”) and the subsequent Ks (subscript “t+1”). The variables d
and s are, respectively, the probabilities of divergence from the identical state and reversion to
the identical state. Since the null model assumes no bias, d and s are insensitive to whether the
identical state is cycling or non-cycling.
Solving equations (3) and (4) for d, we obtain:

d  1

Dt s Ct 1

2Ct
Ct

57

(6)

d  1

Dt s N t 1

2Nt
Nt

(7)

Using the property that the right hand sides of (6) and (7) must be equal, we arrive at the
following formula for s that depends solely on duplicate frequencies:
2Ct 1 2 N t 1

Ct Dt N t Dt
s
1
1

Ct N t

(8)

Initial conditions were set equal to values of C, D, and N observed at Ks = 0.3. We first
attempted to fit values for d and s using frequencies at Ks 0.3 and 0.6, however because the
percentage change in C is greater than N, we obtained a negative value for s. Since s is a
probability, this results is unrealistic, so instead we fit d and s using Ks 0.6 and 0.9, obtaining
values for d (0.42) and s (0.53) that were within [0,1]. Using these parameters, our model was
able to replicate the overall behavior we observed, including the initial dip in C, though the
percentage change is less than that of N (Supplemental Figure 2.3). The root mean squared
error between our predictions and observation was 0.03.

58

Supplemental Figure 2.1 Period, amplitude, and phase of cyclic expression amongst
predictions made by COSPOT, DFT, and both methods combined. (A) The distribution of
59

Supplemental Figure 2.1 (cont’d)
the period of expression in cycling genes predicted by COSPOT (red), DFT (yellow) and both
methods combined (blue). (B) Mean expression (x-axis) vs. the amplitude of cyclic expression
(y-axis) of cycling genes. Color labels follow (A). (C) Phase of expression of cycling genes.
Color labels follow (A).

60

Supplemental Figure 2.2 Most over- and under-enriched GO terms amongst phase clusters
of cycling genes. (A) Heatmap showing the –log10 transformed Fisher’s exact test p-values
(pval) of the top five GO terms with over-represented numbers of genes in each phase cluster
61

Supplemental Figure 2.2 (cont’d)
(ZT 0, 3, 6, 9, 12, 15, 18, and 21). (B) Heatmap showing transformed p-values of GO terms with
under-represented numbers of genes in at least one phase cluster (same as in (A)). P-values were
calculated and transformed as in part (A) except that the left-tail p-value was used.

62

Supplemental Figure 2.3 Divergence of duplicate gene expression state modeled as a system
of difference equations. The frequency at which duplicate pairs in C. reinhardtii are both
cycling (blue), both non-cycling (red), or divergent expression (orange) as a function of the
synonymous substitution rate (Ks). The difference equations used to generate these data are
described in the Supporting Information.

63

Supplemental Figure 2.4 Precision-recall and AUC-ROC curves of SVM predictions for C.
reinhardtii. (A) Precision-recall curves for the prediction of the each of the eight phase clusters
in cycling genes in C. reinhardtii as classified using SVM. Each phase-cluster is represented as a
different colored line: 0 (red), 3 (orange), 6 (lime), 9 (green), 12 (teal), 15 (blue), 18 (purple), 21
(pink). Error bars represent the variance in 10 separate runs of the SVM classifier at optimal
64

Supplemental Figure 2.4 (cont’d)
parameters. (B) ROC curves for the prediction of each of the eight phase clusters in cycling
genes in C. reinhardtii as classified using SVM. Line color and error bars are assigned as in (A).

65

Supplemental Figure 2.5 Regression of the AUC-ROC of phase-expression clusters against
cluster size, and Pearson Correlation Coefficient (PCC) of genes in the cluster. (A) Plot of
phase-expression cluster size against AUC-ROC. The black line indicates the best linear
66

Supplemental Figure 2.5 (cont’d)
regression of AUC-ROC against cluster size. The equation is reported above the figure. (B) Plot
of the mean PCC amongst genes in each phase-expression cluster against AUC-ROC. The black
line indicates the best linear regression of AUC-ROC against mean PCC. The equation is
reported above the figure. (C) Plot of the standard deviation of PCC amongst genes in each
expression cluster against AUC-ROC. The black line indicates the best linear regression of
AUC-ROC against standard deviation of PCC. The equation is reported above the figure.

67

Supplemental Figure 2.6 Expression profiles of cell cycle genes (MAT3, E2F, CDKA1, and
CDKB1) in C. reinhardtii grown in TAP (Tris-Acetate-Phosphate) culture. As observed in
previous studies of C. reinhardtii grown on autotrophic conditions (BISOVA et al. 2005), MAT3,
CDKA1, and CDKB1 are most highly expressed between 12 and 18 hours after dawn, while E2F
expression increases slightly earlier (between 6 and 9 hours).

68

Supplemental Figure 2.7 Distribution of Fourier Transform cyclic score and COSPOT pvalues. Plot of Fourier Transform cyclic score (x-axis) against the negative log transform of the
COSPOT p-value (y-axis). The black line is the best fit power-law regression of the transformed
COSPOT p-value against Fourier Transform cyclic score. The red lines indicated the score
threshold at a significance level of Îą < 0.02 for the Fourier Transform cyclic score (vertical) and
the transformed COSPOT p-value (horizontal).

69

Supplemental Table 2.1 Distribution of Fourier Transform cyclic score and
COSPOT p-values
GO Term

adjusted p-

Description

value1
GO:0019861

4.38E-28

flagellum

GO:0005929

3.07E-11

cilium

GO:0035086

4.41E-06

axoneme

GO:0005874

1.16E-04

microtubule

GO:0007018

4.59E-04

microtubule-based movement

GO:0010287

4.59E-04

regulation of glucose transport

GO:0003777

6.53E-04

microtubule motor activity

GO:0009765

1.19E-03

carbohydrate mediated signaling

GO:0022625

1.19E-03

cytosolic large ribosomal subunit

GO:0006260

1.71E-03

DNA replication

GO:0030286

1.89E-03

dynein complex

GO:0005886

2.16E-03

plasma membrane

GO:0030030

6.12E-03

cell projection organization

GO:0005794

6.52E-03

Golgi apparatus

GO:0042995

1.00E-02

cell projection

GO:0005198

1.00E-02

structural molecule activity

GO:0003774

1.00E-02

motor activity

GO:0022627

1.00E-02

cytosolic small ribosomal subunit

GO:0005774

1.42E-02

vacuolar membrane

GO:0009653

1.77E-02

anatomical structure morphogenesis

GO:0009535

1.77E-02

chloroplast thylakoid membrane

GO:0009637

2.20E-02

response to blue light

GO:0006364

2.20E-02

rRNA processing

GO:0005932

2.65E-02

microtubule basal body

GO:0009506

2.65E-02

plasmodesmata

GO:0004674

2.65E-02

protein serine/threonine kinase
70

Supplemental Table 2.1 (cont’d)
GO:0005509

2.66E-02

calcium ion binding

GO:0005488

2.75E-02

binding

GO:0030992

2.75E-02

intraciliary transport particle B

GO:0009507

3.12E-02

chloroplast

GO:0010114

3.34E-02

response to red light

GO:0010218

3.34E-02

response to far red light

GO:0046686

3.79E-02

response to cadmium ion

GO:0009523

4.16E-02

photosystem II

GO:0048046

4.21E-02

apoplast

GO:0006270

4.21E-02

DNA replication initiation

GO:0009296

4.21E-02

flagellum assembly

GO:0010020

4.21E-02

chloroplast fission

GO:0009434

4.21E-02

motile cilium

GO:0044430

4.21E-02

cytoskeletal part

GO:0019253

4.21E-02

reductive pentose-phosphate cycle

GO:0019773

4.21E-02

proteasome core complex, alpha-subunit
complex

GO:0009826

4.21E-02

unidimensional cell growth

GO:0004298

4.33E-02

threonine-type endopeptidase activity

1. Fisher Exact Test p-value adjusted according to Benjamini-Hochberg

71

Supplemental Table 2.2 Descriptions of the GO terms in each of the five broad functional
categories
Category

GO Terms

Description

photosynthesis and

GO:0015671

oxygen transport

light response

GO:0009773

photosynthetic electron transport in photosystem I

GO:0010206

photosystem II repair

GO:0009765

photosynthesis, light harvesting

GO:0015979

photosynthesis

GO:0010218

response to far red light

GO:0009637

response to blue light

GO:0010114

response to red light

GO:0010304

PSII associated light-harvesting complex II catabolic process

GO:0010020

chloroplast fission

GO:0009507

chloroplast

GO:0009579

thylakoid

GO:0009570

chloroplast stroma

GO:0009941

chloroplast envelope

GO:0009543

chloroplast thylakoid lumen

GO:0009523

photosystem II

GO:0009522

photosystem I

GO:0009533

chloroplast stromal thylakoid

GO:0009534

chloroplast thylakoid

GO:0010287

plastoglobule

GO:0009535

chloroplast thylakoid membrane

GO:0016168

chlorophyll binding

GO:0006260

DNA replication

GO:0006270

DNA replication initiation

GO:0006268

DNA unwinding involved in replication

GO:0000910

cytokinesis

GO:0000724

double-strand break repair via homologous recombination

GO:0006302

double-strand break repair

GO:0007062

sister chromatid cohesion

GO:0007067

mitosis

GO:0006259

DNA metabolic process

GO:0007049

cell cycle

GO:0051726

regulation of cell cycle

cell cycle and mitosis

72

Supplemental Table 2.2 (cont’d)
GO:0006281

DNA repair

GO:0051301

cell division

GO:0006310

DNA recombination

GO:0005819

spindle

GO:0005815

microtubule organizing center

GO:0005694

chromosome

GO:0004003

ATP-dependent DNA helicase activity

GO:0003887

DNA-directed DNA polymerase activity

GO:0004386

helicase activity

microtubules and

GO:0000226

microtubule cytoskeleton organization

flagella

GO:0007018

microtubule-based movement

GO:0009296

flagellum assembly

GO:0030030

cell projection organization

GO:0042384

cilium assembly

GO:0015630

microtubule cytoskeleton

GO:0044430

cytoskeletal part

GO:0019861

flagellum

GO:0030286

dynein complex

GO:0035086

cilium axoneme

GO:0005813

centrosome

GO:0005932

microtubule basal body

GO:0005874

microtubule

GO:0005929

cilium

GO:0005856

cytoskeleton

GO:0005858

axonemal dynein complex

GO:0035085

cilium axoneme

GO:0009434

motile cilium

GO:0030992

intraflagellar transport particle B

GO:0042995

cell projection

GO:0044463

cell projection part

GO:0005876

spindle microtubule

GO:0003777

microtubule motor activity

GO:0003774

motor activity

GO:0004835

tubulin-tyrosine ligase activity

73

Supplemental Table 2.2 (cont’d)
mitochondria and

GO:0006096

glycolysis

metabolism

GO:0006122

mitochondrial electron transport, ubiquinol to cytochrome c

GO:0005983

starch catabolic process

GO:0006098

pentose-phosphate shunt

GO:0007005

mitochondrion organization

GO:0006508

proteolysis

GO:0015986

ATP synthesis coupled proton transport

GO:0045261

proton-transporting ATP synthase complex, catalytic coreF(1)

GO:0005750

mitochondrial respiratory chain complex III

GO:0005747

mitochondrial respiratory chain complex I

GO:0005739

mitochondrion

GO:0005759

mitochondrial matrix

GO:0005741

mitochondrial outer membrane

GO:0005743

mitochondrial inner membrane

GO:0046933

proton-transporting ATP synthase activity, rotational

GO:0046961

mechanism
proton-transporting ATPase activity, rotational mechanism

ribosome and

GO:0006414

translational elongation

translation

GO:0006412

translation

GO:0022626

cytosolic ribosome

GO:0022625

cytosolic large ribosomal subunit

GO:0022627

cytosolic small ribosomal subunit

GO:0019843

rRNA binding

GO:0003735

structural constituent of ribosome

74

Supplemental Table 2.3: Optimal parameters and performance measures of SVM
classification
Phase

C1

R2

AUC-ROC

F-measure

Precision

Recall

0

0.01

4

0.64

0.22

0.24

0.20

3

0.1

1.5

0.62

0.21

0.14

0.39

6

0.1

4

0.62

0.19

0.26

0.15

9

0.1

2.5

0.58

0.22

0.27

0.18

12

0.01

3.5

0.64

0.19

0.21

0.18

15

0.1

1.5

0.64

0.27

0.21

0.40

18

0.1

4

0.65

0.23

0.24

0.23

21

0.01

3.5

0.61

0.21

0.38

0.15

1. C = minimum separation
2. R = ratio of negative to positive examples

75

Supplemental Table 2.4 “Gold Standard” cycling genes in C. reinhardtii
Gene

Name

Reference

Locus

DFT

COSPOT

Cyclic p-value
Score
ATP2/ARF1 ADP-ribosylation
CAH1

MEMON et al.

Cre17.g698000 0.90

1.9e-02

Cre04.g223100 0.76

1.7e-01

JACOBSHAGEN Cre16.g670950 0.78

6.2e-01

factor

(1995)

carbonic anhydrase

FUJIWARA et
al. (1996)

CYC4

cytochrome c

et al. (2001)
Cytosolic

Cytosolic

LEMAIRE et al.

Cre09.g391900 0.29

4.0e-01

thioredoxin

thioredoxin h1

(1999)

chloroplastic

JACOBSHAGEN Cre01.g006950 0.95

1.2e-02

fructose-

et al. (2001)

h1
FBA1

bisphosphate
aldolase
FBA2

chloroplastic

JACOBSHAGEN Cre02.g093450 0.83

fructose-

et al. (2001)

3.4e-02

bisphosphate
aldolase
FBA3

chloroplastic

JACOBSHAGEN Cre05.g234550 0.9

fructose-

et al. (2001)

2.3e-02

bisphosphate
aldolase
FBA4

chloroplastic

JACOBSHAGEN Cre02.g115650 0.61

fructose-

et al. (2001)

bisphosphate
aldolase

76

1.9e-02

Supplemental Tabel 2.4 (cont’d)
FNR1
HSP70B
LCHII
PRK1

Ferredoxin NADP

LEMAIRE et al.

Cre11.g476750 0.81

reductase

(1999)

70kd family heat

JACOBSHAGEN Cre06.250100

2.4e-02

0.32

8.5e-01

shock protein

et al. (2001)

Chlorophyll binding

JACOBSHAGEN Cre06.g283950 0.82

6.5e-03

protein

et al. (1996)

phosphoribulokinase LEMAIRE et al.

Cre12.g554800 0.99

1.0e-02

JACOBSHAGEN Cre12.g542250 0.67

1.1e-02

(1999)
TUB1

Beta-tubulin

& JOHNSON
(1994)
TUB2

Beta-tubulin

JACOBSHAGEN Cre12.g549550 0.98

1.1e-02

& JOHNSON
(1994)
TufA

Elongation factor Tu HWANG et al.
(1996)

77

Cre06.g259150 0.77

1.1e-02

Supplemental Table 2.5 Performance COSPOT and DFT on C. reinhardtii
Genome Coverage1

Gold Stand Coverage2

Îą = 0.01

21.0% (3590)

6.7% (1)

Îą = 0.02

37.4% (6400)

46.7% (7)

Îą = 0.05

54.9% (9392)

73.3% (11)

Îą = 0.01

29.6% (5061)

33.3% (5)

Îą = 0.02

37.6% (6443)

53.3% (8)

Îą = 0.05

55.8% (9556)

73.3% (11)

Method and Îą levels
COSPOT

DFT

1. Parentheses indicated the actual number of genes covered
2. Parentheses indicated how many of 15 gold standard genes are identified as cyclic

78

Supplemental Table 2.6 Performance of combining COPSOT and DFT on C. reinhardtii
Test

Joint

P-value

Probability1

Îą = 0.01

0.0134

Îą = 0.02
Îą = 0.05

Overlap2

Genome

Gold Stand

Coverage3

Coverage4

39.6% (2414)

35.7% (6236)

40.0% (6)

0.0272

56.7% (4579)

47.2% (8072)

73.3% (11)

0.0734

73.5% (8024)

61.7% (10552)

86.6% (13)

1. The joint probability of a gene having a score with a p-value of Îą in either COSPOT or DFT
2. Parentheses indicated the actual number of genes in the overlap set
3. Parentheses indicated the actual number of genes covered
4. Parentheses indicated how many of 15 gold standard genes are identified as cyclic

79

Supplemental File 2.1
Supplemental File 2.1 can be found at the following link:
http://www.g3journal.org/highwire/filestream/472423/field_highwire_adjunct_files/10/FileS3.xl
sx

80

Supplemental File 2.2
Supplemental File 2.2 can be found at the following link:
http://www.g3journal.org/highwire/filestream/472423/field_highwire_adjunct_files/11/FileS4.xl
sx

81

Supplemental File 2.3
Supplemental File 2.3 can be found at the following link:
http://www.g3journal.org/highwire/filestream/472423/field_highwire_adjunct_files/12/FileS5.txt

82

Supplemental File 2.4
Supplemental File 2.4 can be found at the following link:
http://www.g3journal.org/highwire/filestream/472423/field_highwire_adjunct_files/13/FileS6.xl
sx

83

Supplemental File 2.5
Supplemental File 2.5 can be found at the following link:
http://www.g3journal.org/highwire/filestream/472423/field_highwire_adjunct_files/14/FileS7.txt

84

Supplemental File 2.6
Supplemental File 2.6 can be found at the following link:
http://www.g3journal.org/highwire/filestream/472423/field_highwire_adjunct_files/15/FileS8.xl
sx

85

Supplemental File 2.7
Supplemental File 2.7 can be found at the following link:
http://www.g3journal.org/highwire/filestream/472423/field_highwire_adjunct_files/16/FileS9.txt

86

Supplemental File 2.8
Supplemental File 2.8 can be found at the following link:
http://www.g3journal.org/highwire/filestream/472423/field_highwire_adjunct_files/9/FileS2.zip

87

CHAPTER 3: PREDICTING CELL-CYCLE EXPRESSED GENES IDENTIFIES
CANONICAL AND NON-CANONICAL REGULATORS OF TIME-SPECIFIC
EXPRESSION IN SACCHAROMYCES CEREVISIAE

88

ABSTRACT
Gene expression is controlled by regulatory proteins know as transcription factors (TFs).
The collection of all TFs, target genes and their interactions in an organism form a gene
regulatory network (GRN), which can produce complex patterns of expression, such as cycling.
However, identifying which interactions regulate expression in a specific context remains a
challenging task complicated by the existence of multiple approaches to characterize GRNs. To
assess how different methods of defining GRNs capture their regulatory function, we predicted
general and phase-specific cell-cycle expression in Saccharomyces cerevisiae using four
regulatory data sets: chromatin immunoprecipitation (ChIP), TF deletion data (Deletion), protein
binding microarrays (PBMs), and position weight matrices (PWMs). Our results indicate that
data sets with the highest coverage of the S. cerevisiae GRN (ChIP, Deletion and all PWMs)
perform best in predicting cell-cycle expression. Furthermore, prediction performance was
improved by including using TF-TF interactions from feed-forward loops as features as well as
by combining the best predictive features from ChIP and Deletion data. The TFs that were the
best predictors of cell-cycle expression were enriched for known cell-cycle regulators, but TFs
important for predictive models built on ChIP and Deletion data were also enriched for GO
annotations related to invasive growth and metabolism, respectively. Finally, analysis of
important TF-TF interactions suggests that the GRN regulating cell cycle expression is highly
interconnected and clustered around four groups of genes, two of which contain known cellcycle regulators, while the other two contain TFs not previously identified as being involved in
cell-cycle expression.

89

INTRODUCTION
Essential biological processes, from the replication of single cells (Spellman et al. 1998)
to the development of multicellular organisms (Tomancak et al. 2002), are dependent on
complex spatially and temporally specific patterns of gene expression. The regulation of gene
expression during the initiation of transcription depends on both core promoter elements that
interact with the RNA-Polymerase II complex (Juven-Gershon et al. 2008) and accessory
elements known as transcription factors (TFs) that further promote or block the recruitment of
RNA-Polymerase to particular promoter regions (Lelli, Slattery, and Mann 2012; Spitz and
Furlong 2012). TFs bind to short DNA-sequences called cis-regulatory sites, often located in the
upstream promoter region of a gene, though not all of these sites are necessarily occupied at all
times (Lelli, Slattery, and Mann 2012; Spitz and Furlong 2012). TFs do not work in isolation to
regulate gene expression. For example changes in the chromatin state around a promoter can
impact TF binding (M. Li et al. 2015; Benveniste et al. 2014; Miller and Widom 2003). TFs also
interact with other TFs. These interactions can be direct, such as cooperative binding of
regulatory sites (Jolma et al. 2015; Kazemian et al. 2013) or indirect, such as
collaborative/competitive binding to sites (Miller and Widom 2003). There can also be higherorder regulation, where the expression of one TF is regulated by other TFs, such that expression
a gene may depend directly on the TF binding to its promoter and indirectly on the regulation of
that TF. The sum total of direct (TF-target gene) and higher order (TF-TF) interactions
regulating transcription in an organism is referred to as a gene regulatory network or GRN
(Macneil and Walhout 2011). However, because TF binding is dependent on cooperative
binding, cofactors, the chromatin state, and the abundance of the TF under the current conditions
(Spitz and Furlong 2012), these direct and higher order interactions are not static. Therefore,

90

when attempting to understand complex patterns of gene expression, it is important to identify
the relevant interactions.
Although it is understood that the interaction between a TF and the promoter of a target
gene is mediated by cis-regulatory elements, inferring whether a TF binds to a specific promoter
in vivo is complicated by the fact that TFs may recognize multiple distinct nucleotides sequences
(Badis et al. 2009). The promiscuity of TF binding can be addressed by representing the size of
the binding motif and nucleotide preference at different positions of the motif as a position
weight matrix (PWM, Y. Li et al. 2015; Wasserman and Sandelin 2004; Stormo et al. 1982).
PWMs of putative cis-regulatory elements can be identified without experimental evidence of TF
binding by looking for the overrepresentation of DNA sequences in the promoters of coregulated
genes using computational models (Wasserman and Sandelin 2004; Y. Li et al. 2015).
Alternatively, the affinity between TFs and their binding sequence(s) can be assayed in vitro
using protein binding microarrays (Bulyk 2007; Berger and Bulyk 2009) or in vivo by chromatin
immunoprecipitation (ChIP, Buck and Lieb 2004; Furey 2012) or with other emerging
technologies like DapSeq (O’Malley et al. 2016). Binding site information from these in vivo/in
vitro assays can be used to define TF-specific PWMs (de Boer and Hughes 2012), or regulatory
interactions can be identified directly by mapping binding sequences/reads to an annotated
genome (Bailey et al. 2013). Finally, regulatory interactions can be identified by screening for
differentially expressed genes in TF knockouts (Reimand et al. 2010). As there is no single
characterization of a regulatory interaction, it is important to have a method to assess how well a
GRN explains a specific expression pattern.
Before we use regulatory interactions to explain complex expression patterns, we first
must define what we mean by a pattern of expression. Most simply, a gene’s pattern of

91

expression is defined based on the magnitude of expression under a defined set of circumstances,
such as a chosen environment/stress (Zou et al. 2011; Uygun et al. 2017), tissue/body part (Segal
et al. 2008; Chikina et al. 2009), function/biological process (Beer and Tavazoie 2004; Panchy et
al. 2014) or combination thereof (Uygun et al. 2017). Grouping genes based on clear criteria
allows expression regulation to be approached as a classification problem where the presence or
absence of a regulatory interaction is used to predict whether or not a gene exhibits the particular
pattern of expression using a machine learning algorithm. To date, such machine learning
algorithms have been applied to predicting expression in a variety of species, including
Caenorhabditis elegans (Chikina et al. 2009), Arabidopsis thaliana (Zou et al. 2011; Uygun et
al. 2017), and Chlamydomonas reinhardtii (Panchy et al. 2014)). This approach has been applied
successfully to predict complex patterns of expression, such as tissue-specific response to stress
(Uygun et al. 2017), while other modeling methods have been applied to co-expression clusters
based on >100 different environmental conditions (Beer and Tavazoie 2004). Despite these
successes, certain types of patterns remain challenging to predict, such as the specific timing of
expression within a cyclic process (Panchy et al. 2014). Previous studies have explored
improving the performance of machine learning-based predictions of gene expression by
supplementing TF-interaction information with additional information such as DNA accessibility
and cross-species conservation (Uygun et al. 2017) and by including information about
combinatorial rules between TFs that can bind to the same promoter (Zou et al. 2011). However,
it remains to be seen if these approaches are useful for predicting timing of gene expression or
identifying regulatory interactions important for controlling that timing.

92

The cell cycle of budding yeast, Saccharomyces cerevisiae, (reviewed in Bahler 2005) is
an ideal system for studying the regulation of complex expression patterns because the
progression of this process is divided into distinct phases: initial growth (G1-phase), DNA
replication (S-phase), intermediate growth (G2-phase), and cell-division (M-phase). Therefore,
clear patterns of expression can be defined based on when genes reach their peak expression
during the cell cycle, and the expression of genes during the cell cycle has been extensively
characterized in S. cerevisiae (Price et al. 1991; Spellman et al. 1998). Furthermore,
transcriptional regulation is known to play a key role in the control of cyclic expression during
the cell cycle (Futcher 2002, Breeden 2003), and there are multiple data sets defining TF-target
interactions in S. cerevisiae on a genome-wide scale (Harbison et al. 2004, Zhu et al. 2009,
Reimand et al. 2010, de Boer and Hughes 2012). For these reasons, we used the S. cerevisiae
cell-cycle as a model of complex expression in order to study the effect of different approaches
for defining the yeast GRN on our ability to correctly characterize transcriptional regulation. The
TF-target interactions were defined using PWMs, PBMs, ChIP-Chip, or Deletion data, and for
each type of interaction data predictions were made using the same machine learning algorithms
across all cell cycle time points, allowing us to examine the usefulness of each type of data. We
also investigated whether performance could be improved by including TF-TF interactions as
model inputs, applying feature selection algorithms to remove uninformative features, and by
combining TF-interaction information from different data types. Once the best performing model
was identified, Gene Ontology (GO) analysis was used to identify biological functions that are
over- or under-represented in TFs most important for predicting the timing of cell cycle
expression. Finally, we used the most important TF-TF interactions from our models to construct

93

putative GRNs, allowing us to identify subclusters of TFs whose interactions are central to
controlling the timing of expression.

94

RESULTS AND DISCUSSION

Comparing TF-target interactions from multiple regulatory data sets
Although there is a single GRN which describes transcriptional regulation in an
organism, different approaches to defining regulatory interactions may results in inferred GRNs
that have greater or lesser degrees of similarity. In this study, TF-target interactions in S.
cerevisiae was defined using four distinct data sets: TF binding sites inferred from ChIP-Chip
experiments (ChIP-Chip), interactions inferred from changes in expression in deletion mutants
(Deletion), TF binding sites inferred from PWMs (PWM), and TF binding sites inferred from
PBM data (PBM) (Table 3.1). Further details about the processing of each data set can be found
in the Methods section. Because of methodological differences, we would expect to find
differences between GRNs defined using different data sets, both in the total number of
regulatory interactions and the specific relationships between TFs and target genes. The number
of TF-target interactions in the S. cerevisiae GRNs varies from 16,602 in the ChIP-Chip data set
to 78,095 in the PWM data set. This almost 5-fold difference in the number of interactions
identified is not due to differences in the amount of data; each data set includes at least 80 TFs
and 4,701 annotated gene ORFs (Table 3.1). The large difference is driven instead by
differences in the average number of interactions per TF, which varies from 105.6 in the ChIPChip GRN to 558.8 in the PBM GRN. The distribution of TF number per target is positively
skewed for the ChIP-Chip (2.23), Deletion (4.04), and PWM (3.43) GRNs, indicating that most
TFs have fewer interactions than the average value while a few have many more interactions.
The majority of TFs were present in more than one data set (Figure 3.1A); however, the number
of interactions that each TF is involved in is only weakly correlated between the ChIP and

95

Table 3.1 Size and origin of GRNs defined from each data set
Data Set

Transcription
Factors

All Genes

Interactions

Source

ChIP-Chip

152

4701

16,062

ScerTF

Deletion

151

5256

26,757

ScerTF

PWM

230

6536

78,095

YeTFaSCO

Expert PWM

104

4740

9726

YeTFaSCO

PBM

81

4922

45,264

Zhu et al. (2009)

96

Figure 3.1 Coverage of TF and TF-interactions by data set (A) Heatmap of the coverage of S.
cerevisiae TFs in GRNs derived from different data sets. Each row represents a TF and each
column represents the GRN derived from a different data set (ChIP-Chip, Deletion, PWM,
PBM). TFs are sorted according the GRNs they are found in such that TFs belonging to the same
set of GRNs are grouped together. The number of TFs belonging to each group is indicated on
97

Figure 3.1 (cont’d)
the right side of the graph. (B) Heatmap of the percentage of TF-target interactions for each S.
cerevisiae TF belonging to each GRN. Each row represents a TF, and each column represents the
GRN derived from a different data set (ChIP-Chip, Deletion, PWM, PBM). Dark red indicates a
higher percentage of interactions found within a data set, while dark blue indicates a lower
percentage of interactions. TFs are ordered as in (A).

98

Deletion (Pearson’s product moment correlation coefficient (PCC) = 0.092), ChIP and PWM
(PCC = 0.109), and Deletion and PWM (PCC=0.046) datasets.
To further investigate the consistency of inferred TF-target interactions, for each TF we
calculated the percentage of total interactions originating from each data set and grouped them
using hierarchical clustering (Figure 3.1B). Although most TFs are found in > 1 GRN, TFs are
primarily clustered based on the GRN in which they are most prominent, which is not
unexpected given that for the majority (80.5%) of TFs, more than half of their interactions were
identified from a single data type. This pattern held true even when TFs unique to a single data
set were excluded: for 73.6% of TFs found in >1 GRN, more than half of their interactions were
from a single data set. We also looked at the overlap of specific interactions (i.e. the same TF
and target gene) between the different data sets, including a subset of the PWM data set
including only curated binding sites (Figure 3.2). Of the 156,710 TF-target interactions
identified, 89.0% were unique to a single data set, with 40.0% of unique interactions belonging
to the PWM data set. As expected, there was a large overlap between the full PWM data set and
the curated PWM subset, totaling 9,458 interactions or 96.8% of all interactions from the curated
PWMs. However, the degree of overlap between the four main GRNs varied; when TF targets
were chosen at random, ChIP-Chip overlap with Deletion (pv=2.37e-65) and PWM (pv<1e-307)
was higher than expected by random chance, but PWM overlap with Deletion (pv=1.74e-111)
and PBM (pv=1.87e-106) lower (see Methods). The number of overlaps between ChIP-Chip and
PBM (0.057) and Deletion and PBM (0.43) were not significant in either direction
(Supplementary Figure 3.1). This suggests that the ChIP-Chip data set is generally more
similar to the other data sets, while PWMs are more dissimilar. Although, given the low overall

99

Figure 3.2 Overlap in TF-target interactions across data sets. Colors of different regions
indicate different data sets: ChIP-Chip (blue), Deletion (yellow), PWM (orange), Expert PWM
(green), PBM (purple).

100

overlap between data sets, we would still expect models built on each data set to perform
differently.
Predicting timing of expression in the S. cerevisiae cell-cycle using direct regulatory
interactions
Previously, regulatory interactions were used to predicted gene expression in S.
cerevisiae (Beer and Tavazoie, 2004) as well as other species (Chikina et al., 2009; Zou et al.,
2011; Panchy et al., 2014; Uygun et al., 2017). Both general patterns of expression (Beer and
Tavazoie, 2004) and response to specific conditions (Zou et al., 2011) have been accurately
predicted, but distinguishing the phases of cycling expression patterns has proven difficult, even
when the phases are associated with distinct functions and the timing of the cycle is expected to
be strictly controlled (Panchy et al., 2014). Our previous attempt to identify regulators of timed
expression relied primarily on computationally identified putative regulatory interactions
(Panchy et al., 2014). However, we can take advantage of the nearly complete characterization of
regulatory interactions in S. cerevisiae to address this question more directly. Because the TFtarget interactions amongst our four data sets show little overlap, we cannot define a single GRN
for S. cerevisiae. Therefore, we chose to compare the predictive power of TF-target interactions
derived from different data sets and determine which are the most useful for predicting cell-cycle
expressed genes.
To examine cell-cycle expression, we used cell-cycle expressed genes from Spellman et
al. (1998), which are available at the Yeast Cell Cycle Analysis Project (http://genomewww.stanford.edu/cellcycle/). In this study, cell-cycle expressed genes are defined as those
genes whose expression oscillates in a sinusoidal-like fashion over the cell cycle with distinct
minima and maxima. These genes can be clustered into broad categories based on the timing or

101

“phase” of peak expression during the cell-cycle. Spellman et al. identified five such clusters of
71 to 300 genes corresponding to the G1, S, S/G2, G2/M, and M/G1 phases of the cell cycle.
(Supplementary Figure 3.2). While it is known that each phase represents a functionally
distinct period of the cell-cycle, the extent to which regulatory mechanisms are distinct or shared
both within cluster and across all phase clusters is unknown. The coverage of the genes in each
phase cluster by TF-target interactions from different data sets varies between 64% and 100%,
but the average coverage of expression clusters is >70% for all data sets (see Supplementary
Table 3.1), such that we expect the results of any predictor to be generalizable across the entire
cluster.
In order to predict both general and phase-specific expression during the cell cycle, we
used a Support Vector Machine (SVM) algorithm to classify S. cerevisiae genes as being cellcycle expressed or not and, independently, classify genes as being expressed in specific phases of
the cell cycle as defined in Spellman et al. (see Methods for details). The performance of each
classifier was assessed using the Area Under the Curve of the Receiver Operating Characteristic
(AUC-ROC), which ranges from a value of 0.5 for a random classifier to 1.0 for a perfect
classifier. To compare different types of interaction data, we used each of the five sets of TFtarget interactions to independently predict expression. The AUC-ROC values for the bestperforming classifiers generated by each data set are reported in Figure 3.3.
From the distribution of AUC-ROC values, there is an apparent relationship between
performance and both the source of TF-target interactions and the timing of expression during
the cell cycle. We confirmed these relationships by doing analysis of variance (ANOVA) on the
performance of classifiers from each data set (see Methods). There was a significant relationship
between AUC-ROC and data set (pv < 2e-16), expression phase (pv < 2e-16), and the interaction

102

Figure 3.3 Performances of classifiers using TF-target interactions across all data sets.
Heatmap of the AUC-ROC values for SVM models trained each set of cell-cycle expressed
genes (all cell-cycle genes and genes expressed during the G1, S, S/G2, G2/M, or M/G1 phase)
and classified using TF-target interactions derived from each feature set (ChIP-Chip, Deletion,
PWM, Expect PWM, and PBM). The reported AUC-ROC for each classifier is the average
AUC-ROC of 100 data sets composed of a balanced number of positive (cell-cycle genes) and
negative (non-cell-cycle genes) classified using the parameters that maximize performance for
that model (see Methods). Dark red shading indicates an AUC-ROC closer to 1 while dark blue
indicates an AUC-ROC closer to zero.

103

between data and phase (pv < 2e-16). This relationship is not entirely dependent on the number
of feature types, as the performance of the PWM classifier remains is unaffected if we only
include features for TFs present in the ChIP-Chip data set or in Deletion data set
(Supplementary Figure 3.3A). Similarly, if only the most important 150 features defined using
SVM weights (see Methods) are included, so that the total number of features is similar to the
number in the ChIP-Chip and Deletion data sets, the AUC-ROC only declines for G1 and
improves slightly for S-G2 and all cyclic genes. However, if we restrict the ChIP-Chip, Deletion,
and PWM to only TF features present the PBM data set (the one with the fewest TFs), we do see
reduction the performance (Supplementary Figure 3.3B), though ChIP-Chip, Deletion and
PWM still perform better than PBM, even with a reduced feature set. This indicates that, after a
certain threshold, reducing the number of TFs covered by set of TF-target interactions will affect
the ability to predict cyclic expression, though the magnitude of this effect is dependent on how
the TF-target interactions were defined.
Overall, these results indicate that both cell-cycle expression in general and timing of
cell-cycle expression can be predicted using direct regulator interactions, with ChIP-Chip
interactions alone able to predict all clusters except S/G2 with an AUC-ROC > 0.7. While this
suggests that there is significant regulatory information present in TF-target interactions that is
relevant to cell-cycle expression, this information is incomplete given that our classifiers are
imperfect. In particular, no set of TF-target interactions can classify S/G2 expressed genes with
an AUC-ROC > 0.7. One possible explanation for this shortcoming is that this phase bridges the
replicative phase (S) and the second growth phase (G2) of the cell-cycle, and therefore represents
a heterogeneous set of genes with diverse functions and regulatory programs. This hypothesis is
supported by the fact that S/G2 genes are not significantly over-enriched for any Gene Ontology

104

terms (Ashburner et al. 2000; Gene Ontology Consortium 2015) and are only significantly underenriched for genes with mitochondrion (GO:0005739), nucleus (GO:0005643), cytoplasm
(GO:0005737), and RNA binding (GO:0003723) annotations, which are also under-enriched for
all expression clusters expect S-phase. Alternatively, direct regulators alone could be insufficient
to characterize the regulation of genes in this phase cluster as higher-order interaction between
regulators could be involved in regulation of S/G2 expression. With respect to the cell cycle and
gene expression timing in general, the question is what sort of regulatory interactions would we
expect to give rise to expression only at a particular time?
Predicting timing of expression during the S. cerevisiae cell-cycle using feed-forward loops
Given that TF-target interactions produced useful, but imperfect classifiers of cell-cycle
expression, our next step was to identify interactions between TFs that can be used to improve
prediction. Previously, statistical enrichment of TF-binding co-occurring amongst co-expressed
genes has been used to identify regulatory interactions that are useful for prediction (Zou et al.,
2011). However, there is no guarantee that these statistically significant interactions are
biologically important. Instead, we decided to focus specifically on “network motifs”, which are
patterns of regulatory interactions that are enriched in a biological network and thus theorized to
be functionally important (Alon 2007a). In particular, we chose to focus on feed-forward loops
(FFLs). An FFL is a network motif that consists of a primary TF that regulates a secondary TF
and a target gene that is regulated by both the primary and secondary TFs (see Figure 3.4A).
This type of network motif is expected to result in peak expression following a delay after the
expression of the primary TF is induced (Alon 2007a), and is therefore a potential regulatory
mechanism for phase-specific expression in the cell-cycle. Furthermore, FFLs can be used to
compose more complex interactions. For example, negative-feedback loops, which have

105

Figure 3.4 Performance of classifiers using only FFLs across all data sets (A) Representative
feed-forward loops (FFLs) in a GRN. The presence of a regulatory interaction between TF1 and
106

Figure 3.4 (cont’d)
TF2 means that any target gene which is co-regulated by both of these TFs is part of a FFL. For
example, TF1 and TF2 form a FFL with both Tar2 and Ta3, but not Tar1 or Tar4 because they
are not regulated by TF2 and TF1, respectively. (B) Heatmap of AUC-ROC values for SVM
classification models of each cell-cycle expression set (All cell-cycle genes and genes expressed
during the G1, S, S/G2, G2,M, or M/G1 phase) using FFLs derived from each feature set (ChIPChip, Deletion, PWM, Expect PWM, and PBM). The reported AUC-ROC for each classifier is
the average AUC-ROC of 100 data sets composed of a balanced number of positive (cell-cycle
genes) and negative (non-cell-cycle genes) classified using the parameters that maximizes
performance for that model (see Methods). Dark red shading indicates an AUC-ROC closer to 1
while dark blue indicates an AUC-ROC closer to zero.

107

previously been identified as being involved in the regulation of biological oscillations (Bertoli,
Skotheim, and de Bruin 2013; Pett et al. 2016), are composed of two FFLs which identical but
for the direction of the regulatory interaction between the TFs. We can potentially capture
elements of more complicated regulatory pathways by identifying their constituent FFLs.
We defined FFLs in S. cerevisiae using the same four types of regulatory data sets used
to identify TF-target interactions. In order to confirm that FFLs do represent a significantly
enriched network motif in S. cerevisiae GRNs, we calculated the expected number of FFLs based
on the total number of interactions in each GRN and the frequency of TF-TF interactions (see
Methods). We compared these expected values to the actual number of FFLs in each of the five
GRNs and found that in each case, more FFLs were present in the GRN than expected, indicating
FFLs are, in fact, an overrepresented network motif (see Table 3.2). TF-TF interactions alone
are highly correlated with the frequency of TFs (r2 = 0.93) and the total number of TF-TF
interactions (r2 = 0.87) in each data set (see Supplementary Figure 3.4). Given that the
occurrence of TF-TF interactions appears to depend on network size and TF frequency, the
enrichment of FFLs indicate interacting TFs co-regulate the same target genes more often than
expected by random chance.
Given that FFLs are enriched in our GRNs, we built models of cell-cycle expression
using only regulation by FFLs as features. As with TF-target regulations, we treated each GRN
independently because there was little overlap between data sets; 97.6% of FFLs were unique to
one data set and no FFL was common to all data sets (see Supplementary Figure 3.5). Fewer of
the 800 cell-cycle genes defined in Spellman et al. (1998) were targets of an FFL, with three of
the five sets having fewer than 50% of genes covered by a FFL (see Supplementary Table 3.2).
Hence, the models made with FFLs will likely be relevant to only a subset of cell-cycle

108

Table 3.2 Observed and expected number of FFLs in GRNs defined using different data
sets

Data Set

Stdv of
Z-Score3
2
Expected FFLs

Observed FFLs

Mean Expected
FFLs1

3777

811

28.47

104.15

Deletion

13,162

2427

49.26

217.90

PWM

75,514

52,915

230.03

98.24

1700

398

19.94

65.26

67,895

47,371

217.64

94.30

ChIP-Chip

Expert PWM
PBM

1. The mean of FFLs expected in a GRN was determined using the cube of the mean
connectivity of the GRN (see Methods)
2. The standard deviation of FFLs expected in a GRN was determined using the cube of the
mean connectivity of the GRN (see Methods)
3. The z-score reflects the difference between the observed and expected number of FFLs
divided by the standard deviation of the expected number of FFLs (see Methods).

109

expressed genes, but may still be useful for identifying TF-TF interactions important for the
regulation of cell-cyclic expression. Using the same machine learning approach for prediction
and assessment, we found the same overall pattern of performance with FFLs as we did using
direct regulators (Figure 3.4B). Again, the best predictions were from GRNs derived from ChIPChip, Deletion, and all PWMs. However, predictions using ChIP-Chip FFLs had the highest
AUC-ROC values for all phases of expression. ChIP-Chip FFL models also had higher AUCROCs for each phase than those based on direct regulation, though it is important to note that the
ChIP-Chip FFL set had a much lower coverage of cell-cycle expressed genes, 34%, compared to
82% for direct regulators. To test how restricting the set of cell-cycle genes impacts
preformance, we used ChIP-Chip TF-target interactions to predict cell-cycle expression for the
same 34% of cell cycle genes and found performance of predictions was improved
(Supplementary Table 3.3) compared to using all cell-cycle genes (see Figure 3.3). Hence, the
improved performance from FFLs may stem from the subset of cell-cycle genes being used
covered by the ChIP-Chip FFL set being easier to predict using any type of regulatory feature.
Given that models based on ChIP-Chip TF-target interactions predict cell-cycle genes
covered by ChIP-Chip FFLs as well as the FFL model, one might assume the information present
in FFLs is redundant with TF-target interactions. However, it is important to remember that this
subset of cell-cycle genes, which is easier to predict, could not be identified without using FFLs.
For this reason, in spite of their limited coverage of cell-cycle genes, FFLs complement TFtarget regulations, specifically by contributing the classification of the subset that they do predict
well. Additionally, the results of the ANOVA analysis described in the previous section
indicated that the interaction between data type and phase of expression had a significant effect
on the performance of the classifier. Hence, further improvement could be gained not only by

110

including both direct TF-target and FFL interactions, but also by combining interaction across
data sets. However, it is unlikely that all the features from any single data set are relevant to
making accurate predictions, so it is necessary to distinguish between important and unimportant
features before attempting to construct a classifier based on different types of TF-targets and FFL
interactions from multiple data sets.
Using feature importance to merge GRNs and improve prediction of cell-cycle expression
Both the improved performance of FFLs on a subset of cyclic-genes and the effect which
data set has on the performance predicting specific time suggest that a better classifier can be
constructed by combining features and data sets. To do this we focused on interactions identified
from the ChIP-Chip and Deletion data sets because these interactions had better predictive
performance than PBM, PWM and Expert PWM interactions. Furthermore, using ChIP-Chip and
Deletion GRNs are expected to be complementary because they identify interactions using
independent methods: ChIP-Chip interactions represent binding in the absence of a proven
change in expression, while in the Deletion data there is evidence of changes in expression, but
not binding.
In order to merge regulatory information from the ChIP-Chip and Deletion GRNs, we
first identified TF and TF-TF interactions that were important for each of the classifiers based on
SVM weight (see Methods). Features enriched in cell-cycle expressed and non-cell-cycle
expressed genes are differentiated by the sign of the weight: positive weights indicate a feature is
over-enriched in cell-cycle genes while a negative weight indicates a feature is under-enriched in
cell-cycle genes. Because we expect importance to vary across phases in a data set-dependent
fashion, we defined the importance of each feature for each phase-specific classifier based on
ChIP-Chip and Deletion data independently. We used the same criteria for importance across all

111

models, we selected features based on four different percentiles of SVM weight: (1) 10th
percentile of positive weights, (2) 25th percentile of positive weights, (3) 10th percentile of
positive and negative weights, (4) 25th percentile of positive and negative weights (see
Methods). Using this approach allowed us to assess if accurate predictions only require cellcycle associated (i.e. positive weight) features, or if performance depends on exclusionary (i.e.
negative weight) features as well.
Before combining features selected using the above criteria, we first assessed the
predictive power of each subset of TF-target or FFL features, separately and combined, for both
ChIP-Chip (Figure 3.5A) and Deletion (Figure 3.5B) interactions. We found the same overall
pattern of performance as previous classifiers; classifiers built using ChIP-Chip FFL subsets
outperformed classifiers from ChIP-Chip direct interactions across all phases, while the
performance of classifiers using Deletion TF-target interactions and FFLs varied depending on
the phase, with FFL classifiers performing better with S/G2 and G2/M genes like before. For all
subsets consisting either entirely of TF-target regulations or FFLs, the 25th percentile of both
positive and negative SVM weights performed best, except for Deletion FFL predictions for the
S/G2 phase. While this would seem to suggest that more features leads to better performance,
these 25th percentile subsets perform equally well or better than the full data set for both TFtarget interactions and FFLs with a few exceptions (Deletion direct interactions for G2/M and
both ChIP direct interactions and FFLs for G1) (see Figures 3.3 and 3.4B). Similarly, when
combining direct regulators and FFLs, the 10th percentile of positive and negative SVM weights
had the best performance in 75% of cases. These results indicate that we can achieve equal or
improved performance predicting cell-cycle expression using a subset of important features, so
long as both features associated with cell-cycle and non-cell-cycle gene expression are included.

112

Figure 3.5 Performance of classifiers built using important features from ChIP-Chip,
Deletion, and combined ChIP-Chip/Deletion data set. (A) Heatmap of AUC-ROC values for
SVM classification models for each cell-cycle expression set (All cell-cycle genes and genes
expressed during the G1, S, S/G2, G2, M, and M/G1 phases) constructed using a subset of ChIPChip TF-target interactions, FFLs, or both. Subsets of features were defined using the importance
of features (either TFs or TF-TF interactions) as follows: features in the top 10th percentile of
importance (Top 10th), in the top 25th percentile of importance (Top 25th), the top and bottom
10th percentiles of importance (Two-way 10th), and the top and bottom 25th percentiles of
importance (Two-way 25th) (see Methods). The reported AUC-ROC for each classifier is the
average AUC-ROC of 100 data sets composed of a balanced number of positive (cell-cycle
genes) and negative (non-cell-cycle genes) examples classified using parameters that maximize
113

Figure 3.5 (cont’d)
performance for that model (see Methods). Because of the wide range of values, the heatmap is
scaled such that the darkest blue value indicates an AUC-ROC <0.6 and the darkest red color
indicates an AUC-ROC > 0.8. (B) Heatmap of AUC-ROC values for SVM classification models
for each cell-cycle expression set (All cell-cycle genes and genes expressed during the G1, S,
S/G2, G2, M, and M/G1 phases) constructed using a subset of Deletion TF-target interactions,
FFLs, or both. Subsets of genes are defined as in (A). AUC-ROC was calculated and the
heatmap colored as in (A). (C) Heatmap of AUC-ROC values for SVM classification models of
each cell-cycle expression set (All cell-cycle genes and genes expressed during the G1, S, S/G2,
G2, M, and M/G1 phases) constructed using a subset of TF-target interactions, FFLs, or both
from combined ChIP-Chip and Deletion data. Subsets of features are defined as in (A) expect
that only the Two-way 10th and Two-way 25th cutoffs are used and they are applied to both the
ChIP-Chip and Deletion data sets. AUC-ROC was calculated and the heatmap colored as in (A).

114

The final models were built by combining ChIP-Chip and Deletion features, including subsets
with both positive and negative weights. The total number of features in each subset can be
found in Table S4. Although some cell-cycle genes are regulated by a TF-target interaction but
not an FFL, to ensure that the results are comparable, we used all cell-cycles genes covered by at
least one TF-target interaction from either data set to assess the combined ChIP-Chip/Deletion
models. As such, it is not unexpected that the performance of combined models using FFLs was
lower compared with those using TF-target interactions (Figure 3.5C). Nevertheless, the AUCROC of the combined FFL models were > 0.70 for all phase clusters (Figure 3.5C) and, except
for G1, outperformed predictors based on any full set of TF-target interactions (Figure 3.3).
Furthermore, the combined FFL models outperformed combined TF-target interaction models in
predicting cell-cycle gene expression during S/G2. Both 25th percentile models had similar
precision (96.0%), but the FFL model had higher recall of cell-cycle expressed genes (49.3%)
than the TF-target interaction model (44.9%). These two models also correctly identified slightly
different subsets of S/G2 expressed genes, with seven correctly predicted only by the FFL model
and four correctly predicted only by the TF-target interaction model. Hence, it is not surprising
that using both TF-target interactions and FFLs showed the best performance for all phases of
cell cycle expression (Figure 3.5).
Overall, the consistency with which classifiers built using both ChIP-Chip and Deletion
data outperform classifiers built with just one data type indicates the power of using
complementary characterizations of a GRN to predict expression. Furthermore, these combined
models outperform classifiers based on single data sets even though they contain fewer total
features. The performance of features which were found to be important in one of our original
models, both alone and in combination, not only indicates that feature selection can be a

115

powerful tool for improving gene expression predictions, but also that features with high
importance may be enriched in TF and TF-TF interactions that are specific to the control of cellcycle expression. While we would expect that many TFs with high importance are cell-cycle
regulators, we may also use important TF-target and TF-TF interactions to discover novel TF
functions that are associated with cell-cycle regulation.
Functions of TFs important for predicting cell-cycle expression
In our analysis of the ChIP-Chip and Deletion data sets, we found that performance of
classifiers could be maintained while including only the most important features. To test if the
selecting for features important to predicting cell cycle expression identifies true biological
regulations, we asked whether the 10th and 25th percentile of TFs features were enriched in cell
cycle-related genes. Of the 25 TFs that have been annotated as cell-cycle regulators in S.
cerevisiae (GO:0051726), 20 were identified as features important to predicting cell-cycle
expression in either the ChIP-Chip or Deletion data set. For ChIP-Chip, the 10th percentile of the
most important TFs from all phases except M/G1 is enriched for cell-cycle genes, while for the
25th percentile of important TFs, only the features of general classifier (i.e. cell-cycle genes
from all phases) are enriched for cell-cycle genes (Fisher’s Exact Test, Supplemental Table
3.5). The pattern of enrichment was less clear for the Deletion interactions. While important
Deletion TFs from either the 10th or 25th percentile are not enriched for cell-cycle genes, the top
three most important TFs from the general classifier are annotated as cell-cycle genes
(Supplemental Table 3.5). Furthermore, the majority of the 25 cell-cycle annotated TFs are
present in the 25th percentile of important features in at least one phase of cell cycle in both the
ChIP-Chip (14) and Deletion (13) data sets. To summarize these findings, the important features
from our classifiers tend to be associated with the cell-cycle, which suggests our data accurately

116

represents the true S. cerevisiae GRN and that our predictive methodology correctly identified
associations between regulators and expression patterns.
While the set of important TFs identified by our classifiers are enriched for cell-cycle
TFs, these TFs still represent the minority of important TFs. To better understand the functions
of these other important TFs, we looked for additional enriched GO Terms in the 10th and 25th
percentile of important TFs from both the ChIP-Chip and Deletion data sets (Supplemental
Table 3.6). We found 124 GO terms over-represented and 5 under-represented in at least one
feature set, with no terms being over-represented in one set and under-represented in another.
One GO term, mitochondrion (GO:0005739), was under-represented in all four feature sets while
cytoplasm (GO:0005737) and membrane components (GO:0016020 and GO:0016021) were
under-represented in both 25th-percentile feature sets. This was expected given that all features
are TFs. There were 19 GO terms over-represented in all four features sets, including several
generic TF functions (e.g. transcription, DNA-templated, regulation of transcription, DNAtemplated , DNA binding), but also more specific functions including the positive regulation of
transcription in response to variety of stress conditions (e.g. salt, starvation, freezing;
Supplemental Table 3.6). This association is not without precedent, as a previous study found
that cell-cycle genes, particularly those involved in the G1-S phase transition, are needed for
heat-shock response (Jarolim et al. 2013). However, our results indicate a much broader overlap
between cell-cycle regulation and stress response.
The majority of over-enriched GO terms were unique either to ChIP-Chip features (45) or
Deletion features (29). In general, ChIP-Chip TFs were over-represented for terms related to
regulation of growth and phenotype switching while Deletion TFs were over-represented for
terms related to metabolism and the regulation of ribosomes. The full list of GO terms and the

117

features sets they are enriched in can be found in Supplemental Table 3.6. Though there is a
large degree of overlap in the TFs present in the ChIP-Chip and Deletion sets, the difference in
TF-target interactions results in different subsets of these TFs being identified as the most
important, and therefore differential enrichment of gene function amongst the best features of
each set. In particular, the best features derived from ChIP-Chip interactions were over-enriched
for growth related functions, while the features derived from Deletion were enriched for
regulation of metabolism. The distinct functions of important TFs from the ChIP-Chip and
Deletion data supports the hypothesis that the improvement in predictive power from combining
feature sets was due to the distinct, but complementary characterization of gene regulation in S.
cerevisiae.
Finally, we identified GO annotations enriched in TFs important for predicting individual
phases of cell-cycle expression. Because we previously identified GO terms enriched in all
regulators of the cell-cycle, we specifically looked for terms that were not only robust to data set
and importance threshold, but also unique to a single phase of the cell cycle. Out of 274 terms
enriched in at least one phase of the cell cycle, 94 were unique to a single phase (see
Supplemental Table 3.7), but only one was enriched in all four data sets (selenite ion response,
GO:007271, in G2M). An additional 20 unique GO terms were enriched in all ChIP-Chip feature
sets and 4 were in all Deletion feature sets; however there were 60 GO terms whose enrichment
in more than one phase is supported by multiple feature sets. This indicates that the regulation of
expression timing across the cell-cycle involves a certain degree of overlap and that we should
be able to find examples of TFs that are important for multiple phases of cell-cycle expression.
We theorize that such “general” regulators of cell-cycle expression are central to GRNs specific

118

to cell-cycle regulation. To test this hypothesis we made use of the importance of both individual
TF-target and regulatory TF-TF interactions to characterize the structure of cell-cycle regulation.
Identifying regulatory modules for cell-cycle expression
Although looking at the importance and functional enrichment of individual TFs is
important for identifying the factors and processes important to the timing of cell-cycle
expression, ultimately what we want understand is how regulatory interactions play a role in
determining time-specific expression across the cell-cycle. In particular, the prominence of
overlapping enriched functions across the cell-cycle suggests that there may be groups or
“modules” of TFs responsible for regulating multiple phases of expression. This is supported by
the observation that 7.9 and 10.6% of TFs are important for >1 phase at the 10th percentile cutoff
and 32.2% and 30.4% are important for >1 phase at the 25th percentile cutoff for ChIP-Chip and
Deletion interactions, respectively. However, when we hierarchically clustered TF interactions
based on their importance for general and cell-cycle phase specific classifiers (Supplemental
Figure 3.6) we found no large clusters of TFs that could be responsible for regulating expression
at multiple phases. One possible explanation for this could be that, although both positive and
negative importance features were necessary to construct a good predictor of cyclic expression,
using the full range of importance values for all features is confounding. For example, if a
module was important for regulating expression at M/G1 and G1, we would expect that the
importance scores for TFs in that module would be highly correlated during those phases, but
could vary from slightly positive to very negative in the other phases. Thus, different criteria are
required for defining potential regulatory modules.
In order to identify regulatory modules without relying on correlated importance scores,
we used TF-TF interactions to build a network of regulators. To begin, we filtered the set of

119

ChIP-Chip TF-TF interactions by the 10th percentile of importance for predicting cell cycle
genes (Figure 3.6A). We then identified TF-TF interactions that were above the 10th percentile
of importance for one or more phases and found that 61% of all interactions important for
predicting cell-cycle genes were also important for predicting at least one phase cluster and
34.8% were within the top 10th percentile for >1 phases. All but one of the interactions
important for predicting phase-specific expression were concentrated around four groups of
genes (colored regions, Figure 3.6A). Two of these groups, Swi6-Swi4-Mbp1 (red), which is a
regulator of the G1/S phase transition, (Iyer et al. 2001; Wittenberg and Reed 2005; Bean,
Siggia, and Cross 2005) and Fkh1-Fkh2-Ndd1, which is involved in the regulation of S/G2 (G.
Zhu et al. 2000) and G2/M (Koranda et al. 2000) expressed genes, are known regulatory
complexes. Therefore, it is not surprising that interactions amongst these groups are primarily
important for early (G1 through S/G2) and middle (S to G2/M) phases of cell-cycle expression,
respectively. In summary, we were able to identify regulatory modules important for predicting
multiple expression phases that are made up of regulatory complexes known to be important for
cell-cycle progression.
We also found interactions important for multiple phases of cyclic-expression that are not
part of canonical cell-cycle regulatory complexes. For example, the feedback loop between Ste12
and Tec1 was identified in our models as an important regulator of gene expression during S/G2
and M/G1. (purple, Figure 3.6A) Ste12 and Tec1 are known form a complex that shares coregulators with Swi4 and Mbp1 to promote filamentous growth (van der Felden et al. 2014), one
of the functions enriched amongst TFs important for predicting cell cycle expression. However,
neither of these TFs interacts directly with the Swi6-Swi4-Mbp1 complex (van der Felden et al.
2014) nor are they part of the annotated set of cell-cycle regulators. Similarly, interactions

120

Figure 3.6 The cell-cycle expression GRN defined using the 10th percentile of ChIP-Chip
features. (A) A network of ChIP-Chip TF-TF interactions selected from the ChIP-Chip GRN
constructed using the ChIP-Chip FFLs from the top 10th percentile (see Methods) of importance
for predicting all cell-cycle expressed genes. Interactions are further annotated with the stage of
cell-cycle expression (1 = G1, 2 = S, 3 = S/G2, 4 = G2/M, 5 = M/G1) they are important for
predicting (10th percentile of SVM weight in ChIP-Chip models). Four modules with interactions
important for predicting >1 phase of cell-cycle expression are highlighted by color: Swi6-Swi4Mbp1 (red), Fkh2-Fkh1-Ndd1 (green), Ste12 and Tec1 (purple) and Rap1-Msn4-Hap1 (blue).
121

Figure 3.6 (cont’d)
(B) A network of TF-TF interactions from the ChIP-Chip GRN which exists amongst TFs in the
top 10th percentile of importance for predicting all cell-cycle expressed genes using ChIP-Chip
TF-target interactions. Genes are colored as in (A).

122

between Rap1, Hap1, and Msn4 are in the top 10th percentile of important ChIP-Chip TF-TF
features for predicting the M/G1 and G1 phases (blue, Figure 3.6A). However, none of these
TFs are annotated cell-cycle regulators; Rap1 is involved in telomere organization (Guidi et al.
2015; Laporte et al. 2016), Hap1 is an oxygen response regulator (Keng 1992; Ter Linde and
Steensma 2002), and Msn4 is a general stress response regulator (8641288, 8650168).
Of the genes involved in complexes that regulate multiple phases of the cell cycle, but
that are not annotated cell-cycle regulators, only Tec1 is found in the 10th percentile of important
TF features in ChIP-Chip data. Furthermore, Rap1 and Hap1 are not found in any important TF
feature set. Rather their SVM weights are near zero, suggesting that direct regulation by Rap1 or
Hap1 is not significantly associated with either cell-cycle expression or non-cell-cycle
expression. Hence, it is only by looking at interaction between regulators that the importance of
these TFs becomes apparent. In addition, had we only considered interaction amongst the 10th
percentile of TF features with the best performance in ChIP-Chip data (Figure 3.6B), we would
missed the Ste12-Tec1 and Rap1-Hap1-Msn4 modules entirely. Additionally, Fkh1 is not
amongst the 10th percentile of TF features, so part of the Fkh1-Fkh2-Ndd1 module would have
been overlooked as well. The network of all interactions amongst the 10th percentile of TFs also
includes many TF-TF interactions involving Ndd1 and Swi5 that were not found to be important
to predicting cell-cycle expression in our classifiers. The source of TF-interactions is significant
as performing the same analysis on the 10th percentile of TF-TF interactions in the Deletion data
set revealed none of the same modules as in the ChIP-Chip networks (Supplemental Figure
3.7). This includes all of the canonical interactions between Fkh1-Fkh2-Ndd1 as well as the
interaction between Swi6-Mbp1. This illustrates the power of identifying potential TF-TF

123

interactions in a way that is independent of individual TF importance, though the results are
highly dependent on how such interactions are defined.
To further investigate important regulatory interactions, we expanded our network to
include the 25th percentile of TF-TF interactions from ChIP-Chip data (Figure 3.7). In the
resulting network, 38 of 46 TFs (82.6%) formed a single network, while only 0.8% of networks
formed by randomly drawing the same number of interactions from ChIP-Chip data had a similar
or greater degree of interactivity. Comparably, 57 of 67 TFs (85.1%) of the 25th percentile of
TF-TF interactions in the Deletion data set are interconnected, but 28.6% of random networks of
equal size have a similar or greater degree of connectivity. We again identified the interactions of
the expanded ChIP-Chip network using the 25th percentile of importance across expression
phases, which resulted in 89% of the TF-TF interactions being significant in at least one phase,
an increase from 61% in the 10th percentile network, but the frequency of interactions important
for >1 phase remained about the same (35.6%). We should also note that the interaction between
Swi4 and Mcm1, which fell just below the 25th percentile cutoff of importance for predicting
general cell cycle expression, was above the 25th percentile cutoff for all phases except for G1,
making it the only near universal regulatory interaction observed in this study. The majority of
the new interactions important for >1 phase originated from one of the four multi-phase modules
identified in the network built from 10th percentile interactions, while the remainder are
distributed through the rest of the network (Cup9-Yap6, Ino4-Met4, and Met32-Ume6).
Therefore, the modular structure identified in the previous ChIP-Chip network appears to be
robust to the threshold we used to define importance.
Overall, the structure of the GRN built from the ChIP-Chip network indicates the
presence of multiple, broad regulatory modules that interact with each other and with peripheral,

124

Figure 3.7 The cell-cycle expression GRN defined using the 25th percentile of ChIP-Chip
TF-TF interactions. A network of TF-TF interactions selected from the ChIP-Chip GRNs
constructed using ChIP-Chip FFLs from the top 25th percentile (see Methods) of importance for
predicting all cell-cycle expressed genes. Interactions are further annotated with the stage of cell125

Figure 3.7 (cont’d)
cycle expression (1 = G1, 2 = S, 3 = S/G2, 4 = G2/M, 5 = M/G1) they are important for
predicting (25h percentile of SVM weight in ChIP-Chip models). Four modules with interactions
important for predicting >1 phase of cell-cycle expression are highlighted by color: Swi6-Swi4Mbp1 (red), Fkh2-Fkh1-Ndd1 (green), Ste12 and Tec1 (purple) and Rap1-Msn4-Hap1 (blue).

126

phase-specific regulators to control expression timing across the cell-cycle. Importantly, this is
only true of the network built from TF-TF interactions in the ChIP-Chip feature set, while the
network derived from Deletion TF-TF interactions lacks the same modularity). Differences in
network structure are not unexpected, given that interactions derived from the ChIP-Chip data
are inferred using direct binding to target promoters, while those from Deletion data include any
target whose expression is affected by the loss of the TF, whether it binds directly or acts
indirectly through another gene. Hence, we interpret the contrasting results from these data sets
to mean that the direct regulation of cell-cycle expression timing involves the regulatory modules
identified in the ChIP-Chip network, while there is another set of regulators identified in the
Deletion network whose net effect on transcription through both direct and indirect regulatory
interactions is also important for timing of expression during specific phases.

127

CONCLUSIONS
Predicting the expression of genes from their regulatory elements remains a challenging
exercise, but one that can be useful for studying how organisms respond to various stimuli and
how that response is regulated at the molecular level. Here, we have shown that the problem of
predicting complex expression patterns, such as the timing of expression across the cell-cycle, is
tractable using a variety of experimental and computational methods of defining TF-target
interactions. In spite of painting distinctly different pictures of the S. cerevisiae GRN,
interactions inferred from ChIP-Chip, Deletion and PWM data sets were useful for predicting
genes expressed during the cell cycle and for distinguishing between genes expressed at different
phases. In fact, because some cell-cycle genes were only correctly predicted using ChIP-Chip or
Deletion data, integrating interactions from both data sets into a single model improved the
overall accuracy of machine learning models. Furthermore, we found that models were improved
with the addition of TF-TF interactions in the form of FFLs and that a subset of the most
important interactions, combined with a subset of the most important TF-target interactions,
performed better than either the full set of TF-target interactions or FFLs.
By studying the TFs involved in the most important TF-target interactions and FFLs we
were able to infer that these interactions play a biologically significant role in regulating the cellcycle. Using GO analysis, we found that the 10th percentile of important TFs from every phase
except M/G1 were enriched for TFs with cell-cycle annotations. For the M/G1 phase we
identified important TF-TF interactions that involve non-canonical cell-cycle regulators, such as
the regulatory modules Ste12-Tec1 and Rap1-Msn4-Hap1. The Rap1-Msn4-Hap1 module stands
out in that, while these regulators are individually poor predictors of cell-cycle expressions,
interactions between these TFs are among the best predictors of both cell-cycle expression in

128

general and of the M/G1 and G1 phases in particular. Our GO analysis also indicated that TFs
important for predicting cell-cycle expression were enriched for genes associated with
metabolism, invasive growth, and stress responses, which was reflected in the network analysis
as we found that interactions important for >1 phases of cell-cycle expression were clustered
around TFs involved in those processes (Cst6, Ste12-Tect, Rpn4, Rap1-Msn4-Hap1).
Even though our best performing data has nearly complete coverage of the S. cerevisiae
transcriptome, our models do not provide a complete picture of the regulation of cell-cycle
expression. In particular, kinases and the interaction between kinases and TFs are known to play
a key role in regulating the timing of the cell cycle, and FFLs are frequently observed in this TFkinase network (CsikĂĄsz-Nagy et al. 2009). Better characterization of TF binding sites will also
help provide more accurate representation of the GRN regulating expression timing, such as
novel methods of characterizing binding sites that incorporate information about both position
and DNA modification (Csikász-Nagy et al. 2009; O’Malley et al. 2016). Nevertheless, this work
shows that predictive models can provide a framework for identifying both regulators and
regulatory interactions with biological significance to processes of interest. Understanding the
molecular basis of the timing of expression is of interest not only to the cell-cycle, but other
important biological processes, such as response to environmental cues, including acute stresses
like predation and infection as well as cyclical changes in the environment such as light and heat.
Furthermore, the approach described here is not limited to the study of expression timing, but
can also be applied to any expression pattern with discrete phases.

129

MATERIALS AND METHODS

TF-target interaction data and regulatory cite mapping
Data used to infer TF-target interactions in S. cerevisiae were obtained from the
following sources: ChIP-Chip (Harbison et al. 2004) and Deletion (Reimand et al. 2010) data
were downloaded from ScerTF (http://stormo.wustl.edu/ScerTF/), PWMs (de Boer and Hughes
2012) and the expert curated subset of these PWMs were downloaded from YetFaSCO
(http://yetfasco.ccbr.utoronto.ca/), and PBM binding scores were taken from Zhu et al. (see
Supplemental Table 5, (C. Zhu et al. 2009). For ChIP-Chip and Deletion data, the interaction
between TF and their target genes were directly annotated, however, for PWMs and PBMs data
we mapped inferred binding sites to the promoters of genes in S. cerevisiae downloaded from
Yeastract (http://www.yeastract.com/). All position weight matrices were mapped for the PWM
data set, however for PBM data we only used the oligonucleotides in the top 10th percentile of
scores for every TF. This threshold was determined using a pilot study which found that using
the 10th percentile as a cutoff maximized performance of prediction using PBM data. Mapping
was done according to the pipeline previously described in Zou et al. (2011) using a threshold
mapping p-value of 1e-5 to infer a TF-target interaction.
Overlap between TF-target interaction data
To evaluate the significance of the overlap in TF-target interactions between different
GRNs, we compared the observed number of overlaps to what we expected were the genes
regulated by each transcription factor randomized. In detail, for each set of TF-target interactions
we replaced the target gene of each interaction with one that was randomly drawn from the total
set of target genes across all data sets, such that the number of interactions for each TF were

130

preserved. For each randomization of target gene, the number of overlapping features between
each pair of data set was calculated. This process was repeated 1000 time to determine the mean
and standard deviation of overlap between each data set expected under this randomization
regimen. To determine to degree to which our observation differed from the expectation under
this random model, we applied the two-tailed z-test to the differences between the observed
number of overlaps and the distribution of overlaps from the randomized trials.
Expected feed-forward loops in S. cerevisiae regulatory networks
FFLs were defined in each set of TF-target interactions as any pair of TFs with a common
target genes where a TF-target interaction also existed between one TF (the primary TF) and the
other (the secondary TF) which, for clarity, we refer to as a TF-TF interaction. The expected
number of FFLs in each data set was determined according to the method described by Uri Alon
in “An Introduction to Systems Biology” (Chapter 4, 2007b). Briefly, the expected number of
FFLs (NFFL) in a randomly arranged GRN is approximated by the cube of the mean connectivity
(Îť) of the network with a standard deviation equal to the square-root of the mean. Therefore, for
each data set we compared the observed number of FFLs to the expected number of FFLs from a
network with the same number of connections, but with those connections randomly arranged by
defining Îť as the number of TF-target interactions divided by the total number of nodes (TFs +
target genes) and calculating mean the standard deviation as above.
Validating FFLs in cell-cycle expression
FFLs were validated in the context of cell-cycle expression by modeling the regulation
and expression of genes involved in the FFL using a system of ordinary differential equations:

𝛼𝑆
𝑆
∆( ) = (
𝛽𝑆,𝑇
𝑇

𝛽𝑃,𝑆
0 𝑆
)( ) + (
) 𝑓(𝑡)
𝛼𝑇 𝑇
𝛽𝑃,𝑇
131

Where S and T are the expression of the secondary TF and target gene respectively, ∝S and ∝T
are the decay rates of the secondary TF and target gene respectively, and βS,T indicates the
production rate of the target gene dependent on the secondary TF. In the nonhomogeneous term
portion of the equation, βP,S and βP,T are the production rate of the secondary TF and target gene,
respectively, which depend on the primary TF, while f(t) is the expression of the primary TF
over time which is independent of both the secondary TF and the target gene. This system was
solved in Maxima (http://maxima.sourceforge.net/index.html). For each FFL, maximum
likelihood estimation, implemented using the bbmle package in R (https://cran.r-project.org/web/
packages/bbmle/index.html), was used to fit the model parameters to the observed expression of
genes during the cell-cycle as defined by Spellman et al. (1998). Each run was initialized using
the same set of initial conditions and only FFLs for which a reasonable (∝ < 0, βs > 0), noninitial parameters could be fit were kept. Between 80 and 90% of FFLs in each data set passed
this threshold, while only 21% of FFLs built from random TF-TF-target triplets were fit.
Classifying cell-cycle genes using machine learning
Predicting cell-cycle expression and phase of cell-cycle expression was done using the
Support Vector Machine (SVM) algorithm implemented in Weka (Hall et al. 2009). For each
SVM run, the full set of positive instances (either cell-cycle genes or genes expressed at a certain
phase of the cell-cycle) and negative instances (genes in the Spellmen et al. expression data set
which were not cell-cycle expressed) was used to generate 100 balanced (i.e. 1-to-1 ratio of
positive to negative) inputs. Genes were only selected for the input of a SVM run if at least one
interaction feature was involving that gene was present. Features consist of the presence of

132

regulation by a TFs FFL, or a combination of both from one or more regulatory data sets (ChIPChip, Deletion, PWM, Expert-PWM, PBM).
Each balanced input set was further divided into 10-folds for cross validation. SVM uses
the training data to define a linear classifier (i.e. a hyperplane) in the space defined by features,
which is then used to classify positive and negative instances in the test set. Each run was
optimized using a grid search of two parameters: the minimum distance between the positive and
negative groups (C) and the ratio of negative to positive examples in the training set (R).The
tested range of each parameter was as follows: C = (0.01, 0.1, 0.5, 1, 1.5, 2.0) and R = (0.25, 0.5,
1, 1.5, 2, 2.5, 3, 3.5, 4). For each pair of parameters, performance was measured using the AUCROC values averaged across the 100 balanced input sets. For each choice of positive class and
feature set, the pair of grid search parameters which maximized the average AUC-ROC was used
to define the representative model for that predictor and calculate the reported AUC-ROC for
that predictor.
Evaluating the relationship between model performance, class and feature
The effect of the phase (general cell-cycle, G1, S, S/G2, G2/M or M/G1) of expression
being predicted (class) and the data set (ChIP-Chip, Deletion, PWM, Expert PWM or PBM)
from which TF-target interactions were derived (feature) on the performance of each SVM
model was evaluated using analysis of variance (ANOVA). This was done using the “aov”
function in the R statistical language using the following model:
𝑆 = 𝐶+𝐷+𝐶∗𝐷
Where “S” is the representative AUC-ROC score of the SVM model, “C” is a categorical feature
representing the positive-class set (cyclic expression or a specific phase of expression), and “D”
is a categorical feature representing the data set of regulations used.

133

Importance of features to predicting cell-cycle expression
The importance of a feature for each model was determined by rerunning each SVM
model using the best pair of parameters with the options “-i -k” in order to generate an output
files with class and features statistics. From the resulting output file, custom Python scripts were
used to extract the weight value for each of the features used in the linear classifier. Features
were then ordered by their weight to determine importance, such that the feature with the largest
positive value (most strongly associated with the positive class) had the highest rank and the
feature with the largest negative value (most strongly associated with the negative class) had the
lowest rank. Because multiple features often had the same weight value, we defined cutoff scores
for the 10th and 25th percentile conservatively, such that the cutoff for the Xth percentile of
positive features was smallest weight above which includes X% or less of all features and the
Xth percentile of negative features was the largest weight below which includes X% or less of all
features. The effect of this is observed most prominently in the 25th percentile features sets as
ties between feature weights were more common towards the middle of the weight distributions.
GO Analysis
GO annotation for genes in S. cerevisiae were obtained from the Saccharomyces Genome
Database (http://www.yeastgenome.org/download-data/curation). The significance of enrichment
of a particular term in a set of important TF was determined using the Fisher’s Exact Test and
adjusted for multiple-hypothesis testing using the Benjamini-Hochberg method (Benjamini and
Hochberg, 1995).

134

ACKNOWLEDGEMENTS
We thank Christina Azodi and Melissa Lehti-Shiu for their assistance editing this manuscript.

135

APPENDIX

136

Supplemental Figure 3.1 Expected overlaps of TF-target interactions across regulatory data sets.
IQR plots of the expected number of overlapping TF-target interactions between each pair of
GRNs based on randomly drawing TF-target interactions from the total pool of interactions
across all data sets (see Methods). Blue points indicate the observed number of overlaps
between each pair of GRNs.

137

Supplemental Figure 3.2 Expression profiles of genes expressed at specific phases of the
cell-cycle. Expression profiles of genes expressed at each phase of the cell-cycle: G1 (red), S
(yellow), S/G2 (green), G2/M (blue), and M/G1 (purple). Time (x-axis) is expressed in minutes
and, for the purpose of display, the expression (y-axis) of each gene was normalized between 0
138

Supplemental Figure 3.2 (cont’d)
and 1. Each figure shows the mean expression of the phase cluster (dark line) and the range of
values (transparent shading).

139

Supplemental Figure 3.3 Performance of classifier using alternative feature sets. (A)
Heatmap of AUC-ROC values for SVM classification models for each cell-cycle expression set
(all cell-cycle genes and genes expressed during the G1, S, S/G2, G2/M, and M/G1 phases) using
TF-target interactions derived from PWM features filtered using TFs found in the ChIP-Chip
data set, TFs found in the Deletion data set, and the 150 PWMs in the original PWM classifier
with the highest absolute important values. The reported AUC-ROC for each classifier is the
average AUC-ROC of 100 data sets composed of a balanced number of positive (cell-cycle
genes) and negative (non-cell-cycle genes) classified using the parameters that maximize
performance for that model (see Methods). Dark red shading indicates an AUC-ROC closer to 1
140

Supplemental Figure 3.3 (cont’d)
while dark blue indicates an AUC-ROC closer to zero. (B) Heatmap of AUC-ROC values for
SVM classification models for each cell-cycle expression set (all cell-cycle genes and genes
expressed during the G1, S, S/G2, G2/M, and M/G1 phases) using TF-target interactions derived
from the ChIP-Chip, Deletion and PWM data sets filtered using the TFs covered by the PBM
data set. AUR-ROC was calculated and the heatmap colored as described in (A)

141

Supplemental Figure 3.4 Relationship between TF genes and TF-TF interactions. (A) The
relationship between the percent of genes that are TFs (x-axis) in each of the five feature sets
(ChIP-Chip, Deletion, PWM, Expert PWM, and PBM) and the percent of TF-TF interactions in
that data set. Blue data points represent the observed values for each data set, and the black line
is the best fit linear trendline between percent TFs and percent TF-TF interactions. The trendline
equation and associated coefficient of determination are reported below the graph. (B) The
142

Supplemental Figure 3.4 (cont’d)
relationship between total number of interactions (x-axis) in each of the five feature sets (ChIPChip, Deletion, PWM, Expert PWM, and PBM) and the number of TF-TF interactions in that
data set. Blue data points represent the observed values for each data set, and the black line is the
best fit linear trendline between percent TFs and percent TF-TF interactions. The trendline
equation and associated coefficient of determination are reported below the graph

143

Supplemental Figure 3.5 Overlap of FFLs across data sets. Venn-diagram of the number of
overlapping FFLs from different feature sets used to predict cell-cycle expression: ChIP-Chip
(blue), Deletion (yellow), PWM (orange), Expert PWM (green), PBM (purple).

144

Supplemental Figure 3.6 Importance of TF features across classification models. Heatmaps
of the importance, determined by SVM weight, of TF features from ChIP-Chip (top) and
Deletion (bottom) data sets across each classifier of cell-cycle expression (All cell-cycle genes
145

Supplemental Figure 3.6 (cont’d)
and genes expressed during the G1, S, S/G2, G2/M and M/G1 phases). TFs in each heatmap are
ordered by hierarchical clustering, and the resulting dendrogram is depicted on the left side of
each heatmap. Darker red indicates that a TF has a more positive SVM weight (i.e. more
enriched in cell-cycle genes) for a given model while darker blue indicates that a TF has a more
negative SVM weight (i.e. more enriched in non-cell-cycle genes).

146

Supplemental Figure 3.7 The cell-cycle expression GRN defined using the 25th percentile of
Deletion TF-TF interactions. (A) A network of TF-TF interactions selected from the Deletion
147

Supplemental Figure 3.7 (cont’d)
GRNs constructed using FFL features from the top 10th percentile (see Methods) of importance
for predicting all cell-cycle expressed genes. Interactions are annotated with the stage of cellcycle expression (1 = G1, 2 = S, 3 = S/G2, 4 = G2/M, 5 = M/G1) they are important for
predicting (10th percentile of SVM weight in Deletion models). For the purpose of contrast,
elements four modules identified as being important for predicting >1 phase of cell-cycle
expression in the ChIP-Chip GRN are highlighted by color: Swi6-Swi4-Mbp1 r (red), Ste12 and
Tec1 (purple) and Rap1-Msn4-Hap1 (blue). This is done to illustrating the disruption of modules
found in the network of TF-TF interactions selected from the ChIP-Chip GRN. (B) A network of
TF-TF interactions from the Deletion GRN constructed using TF-target interactions in the top
10th percentile of importance for predicting all cell-cycle expressed genes. Genes are colored as
in (A).

148

Supplemental Table 3.1 Coverage of cell-cycle genes by TF-target interactions in each data
set
Expression Cluster

ChIP-Chip1

Deletion1

PWM1

Expert PWM1

PBM1

G1

238 (79%)

242 (81%)

292 (97%)

192 (64%)

250 (83%)

S

59 (83%)

62 (87%)

71 (100%)

51 (72%)

61 (86%)

S/G2

94 (78%)

107 (88%)

117 (97%)

93 (77%)

104 (86%)

G2/M1

163 (84%)

177 (91%)

190 (97%)

148 (76%)

163 (84%)

M1/G

98 (87%)

103 (91%)

112 (99%)

74 (65%)

93 (82%)

Total

652 (82%)

691 (86%)

782 (98%)

558 (70%)

671 (84%)

1. Parentheses indicate the percentage of total genes in the expression cluster covered

149

Supplemental Table 3.2 Coverage of cell-cycle genes by FFL interactions in each data set
Expression

ChIP-Chip1

Deletion1

PWM1

Expert

PBM1

PWM1

Cluster
G1

98 (33%)

125 (42%)

266 (89%)

40 (13%)

185 (62%)

S

20 (28%)

36 (51%)

64 (90%)

15 (21%)

51 (72%)

S/G2

34 (37%)

54 (45%)

112 (93%)

30 (25%)

83 (69%)

G2/M1

72 (39%)

104 (53%)

178 (91%)

41 (21%)

128 (66%)

M1/G

44 (39%)

66 (58%)

99 (88%)

23 (20%)

69 (61%)

Total

268 (34%)

385 (48%)

719 (90%)

149 (19%)

516 (65%)

1. Parentheses indicate the percentage of total genes in the expression cluster covered

150

Supplemental Table 3.3 Performance of classifiers built using TF-target interactions on
only cell-cycle genes covered by ChIP-Chip FFLs

AUC-ROC

All

G1

S

S/G2

G2/M

0.74

0.8

0.77

0.76

0.78

151

M/G1
0.79

Supplemental Table 3.4 Total number of feature present in each model built from
combined features sets
Feature Set

Cyclic G1

S

S/G2

G2/M

M/G1

Direct, Two-way 20%

54

84

51

52

45

41

Direct, Two-way 50%

114

113

111

114

104

93

FFLs, Two-way 20%

113

74

97

68

46

57

FFLs, Two-way 50%

263

217

221

199

126

136

Direct and FFLs, Two-way 20%

166

125

147

119

90

97

Direct and FFLs, Two-way 50%

376

329

331

312

229

228

152

Supplemental Table 3.5 Enrichment of TFs with cell-cycle regulation GO annotation in
features of the ChIP-Chip and Deletion data sets
Feature Set

Cyclic

G1

S

S-G2

G2-M

M-G1

ChIP-Chip, 10th

7.31E-06

0.035

0.0004

0.004

0.0007

0.085

0.0003

0.099

0.099

0.26

0.27

0.1

0.42

0.123

0.41

0.41

1

0.11

1

0.2755

1

0.78

0.27

0.58

Percentile
ChIP-Chip, 25th
Percentile
Deletion, 10th
Percentile
Deletion, 25th
Percentile

153

Supplemental Table 3.6 Over and under enrichment of GO Terms in ChIP-Chip and
Deletion feature sets
GO Term

Ench.1

CC2,

CC2,

D3,

D3,

10th

25th

10th

25th

Description

GO:0009074

Over

No

No

No

Yes

aromatic amino acid family catabolic process

GO:0036003

Over

No

No

No

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to stress

GO:0016458

Over

No

No

No

Yes

gene silencing

GO:1901717

Over

No

No

No

Yes

positive regulation of gamma-aminobutyric acid
catabolic process

GO:0001133

Over

No

No

No

Yes

RNA polymerase II transcription factor activity,
sequence-specific transcription regulatory region
DNA binding

GO:0045848

Over

No

No

No

Yes

positive regulation of nitrogen utilization

GO:0061414

Over

No

No

No

Yes

positive regulation of transcription from RNA
polymerase II promoter by a nonfermentable
carbon source

GO:0061400

Over

No

No

No

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to calcium ion

GO:1901714

Over

No

No

No

Yes

positive regulation of urea catabolic process

GO:0008301

Over

No

No

No

Yes

DNA binding, bending

GO:0050801

Over

No

No

No

Yes

ion homeostasis

GO:0001185

Over

No

No

No

Yes

termination of RNA polymerase I transcription
from promoter for nuclear large rRNA transcript

GO:0000183

Over

No

No

No

Yes

chromatin silencing at rDNA

GO:1900008

Over

No

No

No

Yes

negative regulation of extrachromosomal rDNA
circle accumulation involved in cell aging

154

Supplemental Table 3.6 (cont’d)
GO:0061423

Over

No

No

No

Yes

positive regulation of sodium ion transport by
positive regulation of transcription from RNA
polymerase II promoter

GO:0042991

Over

No

No

Yes

No

transcription factor import into nucleus

GO:0031930

Over

No

No

Yes

No

mitochondria-nucleus signaling pathway

GO:0009410

Over

No

No

Yes

No

response to xenobiotic stimulus

GO:0001228

Over

No

No

Yes

No

transcriptional activator activity, RNA polymerase
II transcription regulatory region sequencespecific binding

GO:0071400

Over

No

No

Yes

No

cellular response to oleic acid

GO:0035957

Over

No

No

Yes

No

positive regulation of starch catabolic process by
positive regulation of transcription from RNA
polymerase II promoter

GO:1900461

Over

No

No

Yes

No

positive regulation of pseudohyphal growth by
positive regulation of transcription from RNA
polymerase II promoter

GO:0000165

Over

No

No

Yes

Yes

MAPK cascade

GO:0031940

Over

No

No

Yes

Yes

positive regulation of chromatin silencing at
telomere

GO:0001085

Over

No

No

Yes

Yes

RNA polymerase II transcription factor binding

GO:0006357

Over

No

No

Yes

Yes

regulation of transcription from RNA polymerase
II promoter

GO:0071468

Over

No

No

Yes

Yes

cellular response to acidic pH

GO:1900399

Over

No

No

Yes

Yes

positive regulation of pyrimidine nucleotide
biosynthetic process

GO:0031335

Over

No

No

Yes

Yes

155

regulation of sulfur amino acid metabolic process

Supplemental Table 3.6 (cont’d)
GO:0006986

Over

No

Yes

No

No

response to unfolded protein

GO:0030968

Over

No

Yes

No

No

endoplasmic reticulum unfolded protein response

GO:1900079

Over

No

Yes

No

No

regulation of arginine biosynthetic process

GO:0033673

Over

No

Yes

No

No

negative regulation of kinase activity

GO:0045821

Over

No

Yes

No

No

positive regulation of glycolytic process

GO:1902352

Over

No

Yes

No

No

negative regulation of filamentous growth of a
population of unicellular organisms in response to
starvation by negative regulation of transcription
from RNA polymerase II promoter

GO:0007124

Over

No

Yes

No

No

pseudohyphal growth

GO:0071470

Over

No

Yes

No

No

cellular response to osmotic stress

GO:0090606

Over

No

Yes

No

No

single-species surface biofilm formation

GO:0010895

Over

No

Yes

No

No

negative regulation of ergosterol biosynthetic
process

GO:0001103

Over

No

Yes

No

No

RNA polymerase II repressing transcription factor
binding

GO:2001278

Over

No

Yes

No

No

positive regulation of leucine biosynthetic process

GO:0071940

Over

No

Yes

No

No

fungal-type cell wall assembly

GO:0000433

Over

No

Yes

No

No

negative regulation of transcription from RNA
polymerase II promoter by glucose

GO:0000430

Over

No

Yes

No

No

regulation of transcription from RNA polymerase
II promoter by glucose

GO:0006525

Over

No

Yes

No

No

arginine metabolic process

GO:1900081

Over

No

Yes

No

No

regulation of arginine catabolic process

GO:0019210

Over

No

Yes

No

No

kinase inhibitor activity

156

Supplemental Table 3.6 (cont’d)
GO:1900464

Over

No

Yes

No

No

negative regulation of cellular hyperosmotic
salinity response by negative regulation of
transcription from RNA polymerase II promoter

GO:0036083

Over

No

Yes

No

Yes

positive regulation of unsaturated fatty acid
biosynthetic process by positive regulation of
transcription from RNA polymerase II promoter

GO:0006990

Over

No

Yes

No

Yes

positive regulation of transcription from RNA
polymerase II promoter involved in unfolded
protein response

GO:0003700

Over

No

Yes

No

Yes

transcription factor activity, sequence-specific
DNA binding

GO:0016602

Over

No

Yes

Yes

No

CCAAT-binding factor complex

GO:0043457

Over

No

Yes

Yes

No

regulation of cellular respiration

GO:0000436

Over

No

Yes

Yes

No

carbon catabolite activation of transcription from
RNA polymerase II promoter

GO:0000982

Over

No

Yes

Yes

Yes

transcription factor activity, RNA polymerase II
core promoter proximal region sequence-specific
binding

GO:0061408

Over

No

Yes

Yes

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to heat stress

GO:0000977

Over

No

Yes

Yes

Yes

RNA polymerase II regulatory region sequencespecific DNA binding

GO:0046983

Over

No

Yes

Yes

Yes

protein dimerization activity

GO:0097239

Over

No

Yes

Yes

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to
methylglyoxal

157

Supplemental Table 3.6 (cont’d)
GO:0010688

Over

Yes

No

No

No

negative regulation of ribosomal protein gene
transcription from RNA polymerase II promoter

GO:0045835

Over

Yes

No

No

No

negative regulation of meiotic nuclear division

GO:0032545

Over

Yes

No

No

No

CURI complex

GO:0051038

Over

Yes

No

No

No

negative regulation of transcription involved in
meiotic cell cycle

GO:0090294

Over

Yes

No

No

No

nitrogen catabolite activation of transcription

GO:0000217

Over

Yes

No

No

No

DNA secondary structure binding

GO:0071406

Over

Yes

No

No

No

cellular response to methylmercury

GO:0043631

Over

Yes

No

No

No

RNA polyadenylation

GO:0046685

Over

Yes

No

No

No

response to arsenic-containing substance

GO:1990526

Over

Yes

No

No

No

Ste12p-Dig1p-Dig2p complex

GO:1990527

Over

Yes

No

No

No

Tec1p-Ste12p-Dig1p complex

GO:0001046

Over

Yes

No

No

No

core promoter sequence-specific DNA binding

GO:2000221

Over

Yes

No

No

No

negative regulation of pseudohyphal growth

GO:0061402

Over

Yes

No

No

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to acidic pH

GO:0001080

Over

Yes

No

No

Yes

nitrogen catabolite activation of transcription from
RNA polymerase II promoter

GO:0090180

Over

Yes

No

Yes

No

positive regulation of thiamine biosynthetic
process

GO:0061410

Over

Yes

No

Yes

No

positive regulation of transcription from RNA
polymerase II promoter in response to ethanol

GO:0061411

Over

Yes

No

Yes

No

positive regulation of transcription from RNA
polymerase II promoter in response to cold

158

Supplemental Table 3.6 (cont’d)
GO:0061401

Over

Yes

No

Yes

No

positive regulation of transcription from RNA
polymerase II promoter in response to a hypotonic
environment

GO:0061407

Over

Yes

No

Yes

No

positive regulation of transcription from RNA
polymerase II promoter in response to hydrogen
peroxide

GO:0001324

Over

Yes

No

Yes

No

age-dependent response to oxidative stress
involved in chronological cell aging

GO:0097236

Over

Yes

No

Yes

No

positive regulation of transcription from RNA
polymerase II promoter in response to zinc ion
starvation

GO:0061412

Over

Yes

No

Yes

No

positive regulation of transcription from RNA
polymerase II promoter in response to amino acid
starvation

GO:0061422

Over

Yes

No

Yes

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to alkaline pH

GO:0061429

Over

Yes

No

Yes

Yes

positive regulation of transcription from RNA
polymerase II promoter by oleic acid

GO:0005667

Over

Yes

No

Yes

Yes

transcription factor complex

GO:0032000

Over

Yes

No

Yes

Yes

positive regulation of fatty acid beta-oxidation

GO:0030154

Over

Yes

No

Yes

Yes

cell differentiation

GO:0089716

Over

Yes

No

Yes

Yes

Pip2-Oaf1 complex

GO:0001078

Over

Yes

Yes

No

No

transcriptional repressor activity, RNA
polymerase II core promoter proximal region
sequence-specific binding

159

Supplemental Table 3.6 (cont’d)
GO:0061395

Over

Yes

Yes

No

No

positive regulation of transcription from RNA
polymerase II promoter in response to arseniccontaining substance

GO:0061426

Over

Yes

Yes

No

No

positive regulation of sulfite transport by positive
regulation of transcription from RNA polymerase
II promoter

GO:0005641

Over

Yes

Yes

No

No

nuclear envelope lumen

GO:1900436

Over

Yes

Yes

No

No

positive regulation of filamentous growth of a
population of unicellular organisms in response to
starvation

GO:2000218

Over

Yes

Yes

No

No

negative regulation of invasive growth in response
to glucose limitation

GO:0001225

Over

Yes

Yes

No

No

RNA polymerase II transcription coactivator
binding

GO:0001226

Over

Yes

Yes

No

No

RNA polymerase II transcription corepressor
binding

GO:0097201

Over

Yes

Yes

No

No

negative regulation of transcription from RNA
polymerase II promoter in response to stress

GO:1900240

Over

Yes

Yes

No

No

negative regulation of phenotypic switching

GO:0060963

Over

Yes

Yes

No

No

positive regulation of ribosomal protein gene
transcription from RNA polymerase II promoter

GO:0071931

Over

Yes

Yes

No

No

positive regulation of transcription involved in
G1/S transition of mitotic cell cycle

GO:0001076

Over

Yes

Yes

No

Yes

transcription factor activity, RNA polymerase II
transcription factor binding

GO:0000790

Over

Yes

Yes

No

Yes

160

nuclear chromatin

Supplemental Table 3.6 (cont’d)
GO:0003676

Over

Yes

Yes

No

Yes

nucleic acid binding

GO:0071483

Over

Yes

Yes

No

Yes

cellular response to blue light

GO:0000122

Over

Yes

Yes

No

Yes

negative regulation of transcription from RNA
polymerase II promoter

GO:0036095

Over

Yes

Yes

Yes

No

positive regulation of invasive growth in response
to glucose limitation by positive regulation of
transcription from RNA polymerase II promoter

GO:0001077

Over

Yes

Yes

Yes

Yes

transcriptional activator activity, RNA polymerase
II core promoter proximal region sequencespecific binding

GO:0008270

Over

Yes

Yes

Yes

Yes

zinc ion binding

GO:0061434

Over

Yes

Yes

Yes

Yes

regulation of replicative cell aging by regulation
of transcription from RNA polymerase II
promoter in response to caloric restriction

GO:0046872

Over

Yes

Yes

Yes

Yes

metal ion binding

GO:0000981

Over

Yes

Yes

Yes

Yes

RNA polymerase II transcription factor activity,
sequence-specific DNA binding

GO:0003677

Over

Yes

Yes

Yes

Yes

DNA binding

GO:0006366

Over

Yes

Yes

Yes

Yes

transcription from RNA polymerase II promoter

GO:0000987

Over

Yes

Yes

Yes

Yes

core promoter proximal region sequence-specific
DNA binding

GO:0061409

Over

Yes

Yes

Yes

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to freezing

GO:0061403

Over

Yes

Yes

Yes

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to nitrosative
stress

161

Supplemental Table 3.6 (cont’d)
GO:0061406

Over

Yes

Yes

Yes

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to glucose
starvation

GO:0061405

Over

Yes

Yes

Yes

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to hydrostatic
pressure

GO:0061404

Over

Yes

Yes

Yes

Yes

positive regulation of transcription from RNA
polymerase II promoter in response to increased
salt

GO:0006355

Over

Yes

Yes

Yes

Yes

regulation of transcription, DNA-templated

GO:0043565

Over

Yes

Yes

Yes

Yes

sequence-specific DNA binding

GO:0045944

Over

Yes

Yes

Yes

Yes

positive regulation of transcription from RNA
polymerase II promoter

GO:0006351

Over

Yes

Yes

Yes

Yes

transcription, DNA-templated

GO:0000978

Over

Yes

Yes

Yes

Yes

RNA polymerase II core promoter proximal
region sequence-specific DNA binding

GO:0005524

Under

No

Yes

No

No

ATP binding

GO:0016021

Under

No

Yes

No

Yes

integral component of membrane

GO:0016020

Under

No

Yes

No

Yes

membrane

GO:0005737

Under

No

Yes

No

Yes

cytoplasm

GO:0005739

Under

Yes

Yes

Yes

Yes

mitochondrion

1. Direction of enrichment
2. CC = ChIP-Chip
3. D = Deletion

162

Supplemental Table 3.7 Over enrichment of GO Terms in ChIP-Chip and Deletion feature
sets for specific phases of cell cycle expression
Term

Unique1

ChIP-Chip,

ChIP-Chip,

Deletion, 10th

Deletion, 25th

10th Percentile

25th Percentile

Percentile

Percentile

GO:0071475

G1

NA

G1

NA

G1

GO:0006363

NA

NA

G1

NA

G1

GO:0061426

G1

G1

NA

NA

G1

GO:0031065

G1

NA

NA

NA

G1

GO:0071280

G1

NA

NA

NA

G1

GO:0045732

G1

NA

NA

NA

G1

GO:0001202

G1

NA

NA

NA

G1

GO:0005635

NA

G1

NA

NA

G1

GO:0003682

NA

G1

NA

NA

G1

GO:0072715

G2M

G2M

G2M

G2M

G2M

GO:0036086

G2M

NA

G2M

G2M

G2M

GO:0043388

NA

NA

G2M

G2M

G2M

GO:2000185

G2M

NA

G2M

NA

G2M

GO:0032048

G2M

NA

G2M

NA

G2M

GO:0000435

NA

NA

G2M

NA

G2M

GO:0033309

NA

G2M

NA

G2M

G2M

GO:0042538

NA

NA

NA

G2M

G2M

GO:0001185

NA

NA

NA

G2M

G2M

GO:0071483

NA

NA

NA

G2M

G2M

GO:0010845

NA

NA

NA

G2M

G2M

GO:1900008

NA

NA

NA

G2M

G2M

GO:0051300

NA

NA

NA

G2M

G2M

GO:0005641

G2M

NA

NA

NA

G2M

163

Supplemental Table 3.7 (cont’d)
GO:0000217

G2M

NA

NA

NA

G2M

GO:0043631

G2M

NA

NA

NA

G2M

GO:0061393

NA

G2M

NA

NA

G2M

GO:0046686

MG1

NA

MG1

NA

MG1

GO:0043433

NA

NA

MG1

NA

MG1

GO:0071276

NA

NA

MG1

NA

MG1

GO:0051457

NA

NA

MG1

NA

MG1

GO:1901717

MG1

MG1

NA

MG1

MG1

GO:1901714

MG1

MG1

NA

MG1

MG1

GO:0045848

MG1

MG1

NA

MG1

MG1

GO:0061415

NA

NA

NA

MG1

MG1

GO:0036003

NA

NA

NA

MG1

MG1

GO:0071466

NA

NA

NA

MG1

MG1

GO:0035948

NA

NA

NA

MG1

MG1

GO:1900079

MG1

MG1

NA

NA

MG1

GO:0034644

MG1

MG1

NA

NA

MG1

GO:0045471

MG1

MG1

NA

NA

MG1

GO:0010768

MG1

MG1

NA

NA

MG1

GO:0006525

MG1

MG1

NA

NA

MG1

GO:0000430

MG1

MG1

NA

NA

MG1

GO:1900081

MG1

MG1

NA

NA

MG1

GO:0031335

MG1

MG1

NA

NA

MG1

GO:0010038

MG1

NA

NA

NA

MG1

GO:0001128

MG1

NA

NA

NA

MG1

GO:0019413

MG1

NA

NA

NA

MG1

GO:0070211

NA

MG1

NA

NA

MG1

164

Supplemental Table 3.7 (cont’d)
GO:0090282

NA

MG1

NA

NA

MG1

GO:0008652

NA

MG1

NA

NA

MG1

GO:0007624

NA

NA

S

S

S

GO:0010512

NA

NA

S

S

S

NA

S

NA

NA

S

GO:0003713

NA

S

NA

NA

S

GO:0071072

NA

S

NA

NA

S

GO:0009062

NA

S

NA

NA

S

GO:0070544

NA

S

NA

NA

S

GO:0035952

NA

S

NA

NA

S

GO:0046020

S

NA

NA

NA

S

GO:0030447

S

NA

NA

NA

S

GO:1900375

NA

NA

NA

S

S

GO:0000156

NA

NA

NA

S

S

GO:0071454

NA

NA

NA

S

S

GO:0071469

NA

NA

NA

S

S

GO:0007126

NA

NA

NA

S

S

GO:0001198

NA

NA

NA

S

S

GO:1900466

NA

NA

NA

S

S

GO:0046685

NA

NA

S

NA

S

GO:0031936

SG2

SG2

NA

NA

SG2

GO:0061425

SG2

SG2

NA

NA

SG2

GO:0000228

SG2

SG2

NA

NA

SG2

GO:0001094

SG2

SG2

NA

NA

SG2

GO:0001093

SG2

SG2

NA

NA

SG2

GO:0030466

SG2

SG2

NA

NA

SG2

165

Supplemental Table 3.7 (cont’d)
GO:0097235

SG2

SG2

NA

NA

SG2

GO:0061424

SG2

SG2

NA

NA

SG2

GO:0010944

NA

SG2

NA

NA

SG2

GO:0032436

SG2

NA

NA

NA

SG2

GO:0006282

SG2

NA

NA

NA

SG2

GO:0001132

SG2

NA

NA

NA

SG2

GO:0010833

SG2

NA

NA

NA

SG2

GO:0071169

SG2

NA

NA

NA

SG2

GO:0031492

SG2

NA

NA

NA

SG2

GO:0051880

SG2

NA

NA

NA

SG2

GO:0071930

NA

NA

NA

SG2

SG2

GO:0061407

NA

NA

SG2

NA

SG2

GO:0009083

NA

NA

SG2

NA

SG2

GO:0061412

NA

NA

SG2

NA

SG2

GO:0001324

NA

NA

SG2

NA

SG2

GO:0043618

NA

NA

SG2

NA

SG2

GO:0071244

NA

NA

SG2

NA

SG2

GO:0006560

NA

NA

SG2

NA

SG2

GO:0061410

SG2

NA

SG2

NA

SG2

GO:1900464

G1

G1,S

G1

G1,S

NA

GO:0061435

G1,G2M,MG1,S

G1,G2M,MG1,S

G2M

G2M

NA

GO:0036083

G2M,MG1,SG2

G2M,MG1,SG2

G2M,SG2

G2M

NA

GO:0001077

G1,G2M,S,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

NA

,SG2

,SG2

,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

,SG2

,SG2

,SG2

GO:0006366

G1,G2M,S,SG2

166

NA

Supplemental Table 3.7 (cont’d)
GO:0001080

G1,MG1

G1,MG1,SG2

MG1,SG2

G1,G2M,MG1,S

NA

G2
G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

,SG2

G2

,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

G2M,MG1,S,SG

G1,G2M,MG1,S

,SG2

,SG2

2

,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

,SG2

,SG2

,SG2

,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

,SG2

,SG2

,SG2

,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

,SG2

,SG2

,SG2

,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

,SG2

,SG2

,SG2

,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

,SG2

,SG2

,SG2

,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

,SG2

,SG2

,SG2

,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

,SG2

,SG2

,SG2

,SG2

GO:0060196

S,SG2

NA

G1

G1

NA

GO:0000977

NA

G1,G2M,SG2

G1

G1,G2M,MG1,S

NA

GO:0046872

GO:0003700

GO:0003677

GO:0000978

GO:0000981

GO:0006355

GO:0043565

GO:0045944

GO:0006351

G1,MG1

NA

NA

NA

NA

NA

NA

NA

NA

NA

G2
GO:0001078

G2M,MG1

G2M,MG1,S

G1

G1,S

NA

GO:0003676

G2M,MG1

G1,G2M,S

G1

G1,S,SG2

NA

GO:0089716

SG2

NA

G1,G2M

G1

NA

167

Supplemental Table 3.7 (cont’d)
GO:0031930

NA

NA

G1,MG1

G1

NA

GO:0010674

NA

NA

G1,MG1

G1,G2M

NA

GO:0061402

NA

NA

G1,SG2

G1,G2M

NA

GO:0009074

NA

G1,MG1

G1,SG2

G1,G2M,S

NA

GO:0031940

NA

NA

G1,G2M

G1,G2M

NA

GO:0032000

SG2

NA

G1,G2M

G1,G2M

NA

GO:0035969

S

S

G1,G2M,MG1

G1,G2M,MG1

NA

GO:1902352

S

S,SG2

G1,G2M,MG1

G1,G2M,MG1

NA

GO:0008270

NA

G1,G2M,MG1,S

G1,G2M,MG1,S

G1,G2M,MG1,S

NA

,SG2

G2

,SG2

GO:1900475

NA

G2M

G1,MG1

G1,G2M,MG1

NA

GO:1900476

NA

G2M

G1,MG1

G1,G2M,MG1

NA

GO:1900472

NA

G2M

G1,MG1

G1,G2M,MG1

NA

GO:0009847

NA

G2M

G1,MG1

G1,G2M,MG1

NA

GO:1900471

NA

G2M

G1,MG1

G1,G2M,MG1

NA

GO:0001081

NA

G2M

G1,MG1

G1,G2M,MG1

NA

GO:1900525

NA

G2M

G1,MG1

G1,G2M,MG1

NA

GO:0001103

S

G1,G2M,S

G1,MG1

G1,G2M,MG1,S

NA

,SG2
GO:0046983

G2M

G1,G2M,MG1,S

G1,S,SG2

,SG2

G1,G2M,MG1,S

NA

,SG2

GO:0097236

NA

NA

G1,SG2

G1,G2M,S,SG2

NA

GO:0097239

NA

G1,G2M

G1,SG2

G1,SG2

NA

GO:0061404

NA

NA

G1,SG2

G1,SG2

NA

GO:1902353

S

S

G1,G2M,MG1

G2M

NA

168

Supplemental Table 3.7 (cont’d)
GO:0005739

MG1

G1,G2M,MG1,S

G2M

G1,G2M,S,SG2

NA

,SG2
GO:0034728

G1,G2M,MG1,S

MG1,S

G2M

G2M

NA

GO:2000679

NA

MG1

G2M

G2M

NA

GO:0010723

NA

MG1,S

G2M

G2M

NA

GO:0046324

NA

SG2

G2M

G2M

NA

GO:0001085

NA

G1,SG2

G2M,MG1

G1,G2M,MG1,S

NA

G2
GO:0010527

NA

NA

G2M,S

G2M,S

NA

GO:0045937

NA

S

G2M,S,SG2

G1,G2M,MG1,S

NA

GO:0070210

NA

NA

MG1

G1,MG1

NA

GO:0033673

NA

G1,SG2

MG1

MG1

NA

GO:0019210

NA

G1,SG2

MG1

MG1

NA

GO:0000989

G2M

G2M

G1,MG1,S

G2M,MG1,S

NA

GO:0016458

NA

NA

MG1,SG2

MG1,SG2

NA

GO:0009410

NA

NA

G2M,MG1,S

S

NA

GO:1900399

NA

MG1

S

G2M,MG1,S

NA

GO:0043619

MG1

MG1

S

S

NA

GO:0090575

SG2

S

S,SG2

S

NA

GO:0036095

G2M,MG1

G2M

G1,SG2

SG2

NA

GO:0001010

S,SG2

S

G2M,SG2

SG2

NA

GO:0005737

G1

G1,G2M,MG1,S

SG2

G1,G2M,S,SG2

NA

,SG2
GO:0061408

NA

NA

SG2

G1,G2M,SG2

NA

GO:1900240

G2M,S

G2M,S

SG2

SG2

NA

GO:2000221

MG1,S

NA

SG2

SG2

NA

169

Supplemental Table 3.7 (cont’d)
GO:2001158

NA

G1,G2M

SG2

SG2

NA

GO:0007124

NA

NA

G1

G2M,S

NA

GO:0000165

G2M

NA

G1

NA

NA

GO:1900463

G1

S

G1

S

NA

GO:0090606

G1

S

G1

S

NA

GO:0072363

G1,G2M,MG1,S

G1,S

G1,G2M

NA

NA

,SG2
GO:0010673

NA

NA

G1,MG1

G2M

NA

GO:0007070

NA

NA

G1,MG1

NA

NA

GO:0042991

NA

NA

G1,MG1

NA

NA

GO:0070491

NA

NA

G1,MG1,S

G2M

NA

GO:1900460

G1

NA

G1,MG1,SG2

NA

NA

GO:2000218

G1,G2M,S

S

G1,SG2

NA

NA

GO:0034225

NA

NA

G1,SG2

NA

NA

GO:0035957

NA

NA

G1,SG2

NA

NA

GO:1900461

NA

NA

G1,SG2

NA

NA

GO:0090180

NA

NA

G2M,MG1,SG2

NA

NA

GO:2000222

S,SG2

NA

G2M,S

S

NA

GO:0016036

NA

NA

G2M,S,SG2

NA

NA

GO:0070417

G2M,MG1,SG2

NA

G2M,SG2

NA

NA

GO:0009450

NA

NA

G2M,SG2

NA

NA

GO:0019740

NA

NA

G2M,SG2

NA

NA

GO:0030154

G1

G1

MG1

NA

NA

GO:0001228

G1

G1

MG1

NA

NA

GO:0090295

NA

G1

MG1

NA

NA

170

Supplemental Table 3.7 (cont’d)
GO:0001046

G1,G2M,MG1,S

G1,G2M,MG1,S

MG1

S

NA

,SG2
GO:0005667

G1

G1

MG1,S

G2M

NA

GO:1900462

NA

NA

MG1,SG2

NA

NA

GO:0097201

MG1,SG2

MG1,S,SG2

NA

G1

NA

GO:2001043

S,SG2

S,SG2

NA

G1

NA

GO:0008301

NA

G2M,S

NA

G1

NA

GO:0000790

G1,G2M,MG1,S

G1,G2M,MG1,S

NA

G1,G2M

NA

G2
GO:0045821

NA

G1,G2M,MG1

NA

G1,G2M,MG1

NA

GO:0001135

MG1

MG1

NA

G1,G2M,MG1,S

NA

GO:0006357

NA

G1,MG1

NA

G1,G2M,MG1,S

NA

GO:0000982

MG1,S,SG2

MG1,S,SG2

NA

G1,G2M,MG1,S

NA

,SG2
GO:0001076

NA

G1,G2M,MG1,S

NA

G1,G2M,S,SG2

NA

G2
GO:0006990

NA

G1

NA

G1,G2M,SG2

NA

GO:0016020

NA

G1,MG1

NA

G1,MG1,S

NA

GO:0061432

NA

G2M

NA

G1,S

NA

GO:0061427

NA

G2M

NA

G1,S

NA

GO:1900478

NA

G2M

NA

G1,S

NA

GO:0000122

G1,G2M,MG1,S

G1,G2M,MG1,S

NA

G1,S,SG2

NA

,SG2

,SG2

GO:0000987

MG1,S,SG2

G1,MG1,S,SG2

NA

G1,S,SG2

NA

GO:0016021

NA

G1,MG1,SG2

NA

G1,S,SG2

NA

GO:0001191

G1,S

G1,MG1,S,SG2

NA

G1,SG2

NA

171

Supplemental Table 3.7 (cont’d)
GO:0033169

NA

S

NA

G1,SG2

NA

GO:0032454

NA

S

NA

G1,SG2

NA

GO:0030907

G1,G2M

G2M

NA

G2M

NA

GO:0071931

G1,G2M,MG1

G2M

NA

G2M

NA

GO:0007074

G2M,S

G2M,S

NA

G2M

NA

GO:0000083

G2M,S

G2M,S,SG2

NA

G2M

NA

GO:0006530

S

G2M,SG2

NA

G2M

NA

GO:0004067

NA

G2M,SG2

NA

G2M

NA

GO:0001133

S

G1,G2M,MG1,S

NA

G2M,MG1

NA

GO:0061414

NA

MG1,SG2

NA

G2M,MG1

NA

GO:0070822

G1,G2M

G2M

NA

MG1

NA

GO:0003674

NA

G1,G2M

NA

MG1

NA

GO:1900423

G2M

G2M,SG2

NA

MG1,S

NA

GO:0009063

G1

G1,MG1

NA

NA

NA

GO:0042128

G1

G1,S

NA

NA

NA

GO:0001159

G1

G1,SG2

NA

NA

NA

GO:0060963

G1,G2M,MG1,S

G1,G2M,MG1,S

NA

NA

NA

G1,G2M,MG1,S

G1,G2M,MG1,S

NA

NA

NA

,SG2

,SG2

G1,G2M,MG1,S

G1,G2M,MG1,S

NA

NA

NA

,SG2

,SG2

GO:2001278

G2M

G2M,MG1

NA

NA

NA

GO:0051038

G2M,MG1

MG1

NA

NA

NA

GO:0001190

MG1,SG2

MG1

NA

NA

NA

,SG2
GO:0001225

GO:0001226

172

Supplemental Table 3.7 (cont’d)
GO:0010688

G1,G2M,MG1,S

S

NA

NA

NA

,SG2
GO:0031496

G2M,S

S

NA

NA

NA

GO:0001012

G2M,S

S

NA

NA

NA

GO:0036033

G2M,S

S

NA

NA

NA

GO:0051019

S

MG1,S

NA

NA

NA

GO:0031848

SG2

S,SG2

NA

NA

NA

GO:0032545

G1,G2M,MG1,S

NA

NA

NA

NA

,SG2
GO:0045835

G2M,MG1

NA

NA

NA

NA

GO:0000821

MG1

S

NA

NA

NA

GO:0010895

NA

G1,G2M

NA

NA

NA

GO:0044374

NA

G1,G2M,MG1

NA

NA

NA

GO:0001084

NA

G1,MG1

NA

NA

NA

GO:0010691

NA

G1,MG1

NA

NA

NA

GO:0071322

NA

G2M,S,SG2

NA

NA

NA

GO:0061416

NA

MG1,S

NA

NA

NA

GO:0070187

SG2

S

NA

NA

NA

GO:0000433

G1,G2M,S

G1,G2M,S

NA

S

NA

GO:0000304

MG1

MG1

NA

S

NA

GO:1900436

G2M

NA

NA

S

NA

GO:0031494

NA

G1

NA

S

NA

GO:0001102

NA

G1

NA

S

NA

GO:0001197

NA

G1

NA

S

NA

GO:1900465

NA

G1

NA

S

NA

GO:0061395

NA

MG1

NA

S

NA

173

Supplemental Table 3.7 (cont’d)
GO:0008134

G2M,S

G2M

NA

SG2

NA

GO:0035390

SG2

S,SG2

NA

SG2

NA

GO:0070200

SG2

S,SG2

NA

SG2

NA

GO:0030968

NA

G1

NA

SG2

NA

GO:0016602

G1

G1

S

NA

NA

GO:0043457

G1

G1

S

NA

NA

GO:0000436

G1

G1,MG1

S

NA

NA

GO:0061434

NA

NA

SG2

G1,G2M,S

NA

GO:0061409

NA

NA

SG2

G1,G2M,S

NA

GO:0061403

NA

NA

SG2

G1,G2M,S

NA

GO:0061406

NA

NA

SG2

G1,G2M,S

NA

GO:0061405

NA

NA

SG2

G1,G2M,S

NA

GO:0006338

NA

NA

SG2

G1,G2M,S

NA

GO:0090419

MG1

MG1

SG2

NA

NA

GO:1903468

MG1

MG1

SG2

NA

NA

GO:1990526

G2M,S

S

SG2

NA

NA

GO:0032298

MG1

NA

SG2

NA

NA

GO:0071468

G1

NA

SG2

S

NA

GO:0006572

NA

MG1

SG2

S

NA

GO:0061422

NA

NA

SG2

S

NA

GO:0061401

NA

NA

SG2

S

NA

GO:0061411

NA

NA

SG2

S

NA

GO:1990527

G2M,S

S

G2M,S,SG2

G2M,S

NA

GO:0071400

SG2

SG2

G1,MG1,SG2

G1,SG2

NA

GO:0061429

SG2

SG2

G1,SG2

G1,SG2

NA

1. Unique indicates that a GO term is only enriched in a single phase across data sets

174

CHAPTER 4: EXPRESSION AND REGULATORY ASYMMETRY IS A FEATURE OF
RETAINED TRANSCRIPTION FACTOR DUPLICATES1

1

The work described in this chapter has been submitted for publication:

Nicholas L. Panchy, Christina B. Azodi, Eamon F. Winship, Ronan C. O’Malley, Shin-Han
Shiu (2017) Expression and regulatory asymmetry is a feature of retained transcription factor
duplicates. Submitted

175

ABSTRACT
Transcription factors (TFs) play a key role in regulating plant development and response
to environmental stimuli. While most genes revert to single copy after whole genome duplication
(WGD) event, transcription factors are retained at a significantly higher rate. To assess why TF
duplicates have higher rates of retention relative to other genes, we used Arabidopsis thaliana as
a model and established linear models with expression, sequence, and conservation features to
predict the extent of duplicate retention following WGD events among TFs and 19 groups of
genes with other functions. We found that TFs in particular are retained more often than would
be expected based on the models. Furthermore, the evolution of TF expression patterns and cisregulatory sites favors the partitioning of ancestral states among the resulting duplicates.
However, this is not because TF duplicates tend to subfunctionalize. Instead, one "ancestral" TF
duplicate retains the majority of ancestral expression and cis-regulatory sites, while the “nonancestral” duplicate is enriched for novel regulatory sites. To investigate how this pattern of
biased partitioning has evolved, we modeled the retention of ancestral expression and regulatory
states in duplicate pairs using a system of differential equations. In our best models, TF duplicate
pairs are preferentially maintained in a partitioned state. Our findings suggest that the TF
duplicates with asymmetrically partitioned ancestral states are maintained because one copy
retains ancestral functions while the other, at least in some cases, acquire novel expression
pattern and/or cis-regulatory sites.

176

INTRODUCTION
Plant genomes are replete with paralogous genes derived from a variety of duplication
events and mechanisms (Panchy et al., 2016). Among them, whole genome duplication (WGD)
events are responsible for most extant duplicate genes (Panchy et al. 2016). Two ancient WGD
events took place prior to the divergence of angiosperms (Jiao et al. 2011). Subsequently, more
than a dozen WGD events have occurred across a variety of angiosperm lineages (Lyons et al.
2008; Lee et al. 2013; Myburg et al. 2014; Renny-Byfield et al. 2014; Soltis et al. 2014; Wang et
al. 2014), including three in the lineage leading to Arabidopsis thaliana (Bowers et al. 2003). As
the last known WGD event in the Saccharomyces cerevisiae (Wolfe and Shields 1997; Kellis et
al. 2004) and human (Panopoulou et al. 2003; Dehal and Boore 2005) lineages occurred prior to
the radiation of angiosperms, WGD occurs more frequently in plants relative to other eukaryotic
lineages.
WGD accounts for ~90% of the expansion of TF families across plants lineages (Maere
et al. 2005) and TFs are consistently enriched among WGD duplicates across divergent plant
species ( Lespinet et al. 2002; Shiu et al. 2005; Carretero-Paulet and Fares 2012). In addition,
plant TF duplicates derived from WGD are retained at higher rates than most plant genes with
other functions (Seoighe and Gehring 2004; Shiu et al. 2005). These duplicate TFs contribute
significantly to plant adaption (Lehti-Shiu et al. 2016), agricultural traits (Zhang et al. 2011), and
domestication (Liu et al. 2015). The expansion of several TF families coincides with major
events in the evolution of plants, such as the migration to land and expansion of flowering plants
(De Bodt et al. 2005, Soltis et al. 2008, Weirauch and Hughes 2011). TF duplication is also
central to the evolution of flowering time (Schranz et al. 2002), floral structures (Theissen and
Melzer 2007) and fruit development (Litt and Irish 2003, McCarthy et al. 2015).

177

Because WGD results in duplication of all genes in a genome, the differences in the
degrees of expansion of different gene families (Blanc and Wolfe 2004; Seoighe and Gehring
2004; Hanada et al. 2008; Li et al. 2016) must result from differential rates of gene retention.
Previously, a collection of features including sequence properties (e.g. gene length), biochemical
activities (e.g. expression level), evolutionary characteristics (e.g. substitution rates), and
annotated functions have been used to assess the properties of retained duplicates in general
(Jiang et al. 2013; Moghe et al. 2014). It is still unclear what properties are associated with
retained TFs, how well these properties explain the differences in retention rates between TFs
and other function groups, and how the retained duplicate pairs differ in their properties that may
shed light on the mechanisms of their retention.
In this study, we first assessed how the retention rates of A. thaliana WGD duplicates
differ among TFs, all other genes, and genes in each of 19 other “function groups” of similar size
to TFs. Next, to identify the features contributing to the differences in the percent of retained
duplicates amongst different function groups of genes, we modeled the percent of retained
duplicates as a function of 34 features of genes in three broad categories (expression, sequence,
and conservation) in each function group. In addition, we examined whether the correlations
between a feature and retention status was consistent across function groups or if some groups,
like TFs, deviated from the norm. Furthermore, to assess how the ancestral and extant functions
of duplicate pairs have diverged relative to their ancestral function, we determined how gene
expression and cis-regulatory sites of TF duplicates have likely evolved post WGD by inferring
the ancestral expression and cis-regulatory states of extant TF duplicates. Finally, we modeled
the evolution of TF WGD duplicates as a system of differential equations which tracks the
change in frequency of duplicate pairs retaining the ancestral state in both, one, or neither, to

178

assess whether the partitioning of TF duplicates pairs is maintained by a bias against losing the
ancestral state in the second duplicate copy.

179

RESULTS AND DICUSSION

Retention of duplicate genes in different function groups following WGD
To assess the factors contributing to the differential retention of TF duplicates from
WGD events and duplicates from WGD events involved in other functions, we first quantified
the degree of duplicate retention of A. thaliana WGD duplicates in 20 different function groups.
These function groups include TFs (Jin et al. 2014) and 19 other groups defined based on Gene
Ontology (GO) molecular functions (see Methods). The other functional groups were chosen
based on their larger sizes for comparisons with TFs and their large differences in duplicate
retention (see below). Within each function group, genes were classified as “WGD-duplicates”
(both duplicate copies retained) or “WGD-singletons” (only one copy retained) depending on
whether there were paralogs in corresponding duplicate blocks (Bowers et al. 2003). Because
duplicate retention is expected to differ across different WGD events, duplicate pairs derived
from the , , and  WGD events (Bowers et al. 2003) were analyzed separately. To test the
association between the duplicate retention and membership in a function group, we calculated
the log odds ratio of genes having a retained WGD-duplicate derived from a specific WGD event
for each function group relative to all A. thaliana genes (see Methods). This odds ratio is used to
indicate the degree of duplicate retention. If duplicate retention of a function group is not
significantly different from the rest of the genome, we would expect a log odds ratio ~ 0. A
positive and a negative log odds ratio indicate that a function group contains a proportionally
higher and lower number of retained WGD-duplicates compared to the genome average,
respectively. Among the 20 function groups examined, the log odds ratios were highly

180

heterogeneous and only TFs and protein kinases had significantly higher degrees of retention
than the genome average for all three WGD events (Figure 4.1).
Although both protein kinases and TFs have odds ratio greater than the genome average
across all three WGD events, based on the confidence interval of the odds of retention (Figure
4.1), the log odds ratios of TF retention are even higher for the older (β and γ) duplication events
than those for protein kinases, indicating that on average, the longevity of TF WGD-duplicates is
higher than that of protein kinases. This observation could be an artefact of the ages of these
events, as some duplicates formed due to WGDs would remain in the genome, but are defined as
such because they are not located in recognizable syntenic blocks. This would be particularly
problematic for Îł duplicates, as there are fewer syntenic regions and they are smaller (Bower et
al., 2003). To address this issue, we included A. thaliana paralogs that may be Îł WGD
duplicates based the criteria used in Panchy et al. (2017) and used their synonymous substitution
rate (Ks, see Methods) to assess if TF retention degree for older duplicates is higher than that of
protein kinases. If we were to consider putative paralogs with Ks around the  event (2.7 < Ks <
2.9) as  WGD duplicates, the log odds ratio of retention from the  event would still be
significant for both kinases (0.52, pv = 0.02) and TFs (1.27, pv < 2.2e-16) compared to the rest
of genome and TFs are still retained more frequently from this event (0.67, pv = 0.005)
(Supplemental Figure 4.1). In summary, TFs were retained more frequently post WGD than
most other function groups irrespective of group size or the timing of the event. Compared to
protein kinases, one of the largest gene families in plants (Lehti-Shiu and Shiu, 2012), that also
have significant higher degree of retention in all WGD events, TFs tend to be retained from older
WGDs, suggesting a higher longevity of duplicates.

181

Figure 4.1 Retention of WGD-duplicate genes in A. thaliana. The duplicate gene retention
rates (log odds ratios) within 20 function groups relative to whole genome. Groups are ordered
182

Figure 4.1 (cont’d)
by the odds in the alpha event. Colors represent different WGD duplication events (ι = orange, β
= green, Îł = blue). Bars indicate the 95% confidence interval of the odds of retention. If the
confidence interval does not overlap with zero, this indicates the odd of retaining a duplicate
gene is significantly different than the genome average from that functional group at the 5%

183

Linear model of WGD-duplicate retention across function groups
Amongst function groups, TFs stand out as one of only two that are retained more often
than the genome average consistently across all WGD events. For the rest of the function groups,
the degrees of retention vary above and below the genome average across WGD events. One
possibility is that the degree of retention correlates with gene numbers among functional groups.
However, gene counts and degrees of retention are only very weakly correlated for any WGD
event (r2;  = 0.05,  = 0.16,  = 0.04; Figure 4.2A). Therefore, the reason for the differences in
degree of retention must involve factors beyond gene function, group size, and timing of
duplication. To address why the degrees of retention differs, we examined how sequence,
expression, conservation, and other miscellaneous features (Figure 4.2B, Supplemental Table
4.1) differ among WGD-duplicate and WGD-singleton genes between function groups. We also
asked how well the degree of retention differences between function groups can be explained by
these different features.
To see how well the features we considered could explain the differences in degree of
retention among function groups, we constructed a linear model of the degree of retention for
each WGD event. Here the degree of retention is the odds ratio defined in the previous section. A
subset of the features examined here has previously been shown to be significantly associated
with retention of WGD-duplicates as a whole without considering individual events (Jiang et al.
2013). Here we choose to separate WGD events because there was a large variance in degree of
retention across events and across function groups (Figure 4.1). Thus the features associated
with each event and function group may differ. Consistent with this, the correlations between
degrees of retention and feature values have different signs (black arrows, Figure 4.2B) and
magnitudes (white arrows, Figure 4.2B) depending on the WGD events. Hence, in the next step

184

Figure 4.2 Linear model of the degree of duplicate retention in function groups based on
genes features. (A) Relationships between gene counts and odds of retention of WGD duplicates
185

Figure 4.2 (cont’d)
across functional groups (ι = orange, β = green, γ = blue). The correspondence between group
sizes (numbers of genes) and degrees of retention (odds ratios) was determined using the square
of the Pearson product-moment correlation coefficient (r2, ι = 0.05, β = 0.16, γ = 0.04). (B) A
heatmap of the Pearson product-moment correlation coefficient (PCC) between the values of a
feature across different function groups (rows) and the odds of retention of functions groups
from a particular WGD event (columns, indicated by the symbols ι, β, and γ). Darker red:
stronger positive correlation. Darker blue: stronger negative correlation. Features with different
sign of correlation across WGD events are indicated by black arrows. Features with a large
(≥0.20) difference in PCCs with the same sign are indicated by open arrows. (C) The observed
odds of duplicate retention (x-axis) for each group plotted against the predicted odds of retention
(y-axis) from the best model for each event (ι = orange, β = green, γ = blue). Dotted line:
equality between predicted and observed retention odds. Values from TFs are indicated by a
black arrow while values from protein kinases are indicated by an open arrow. Red dot (TFÎł'):
the predicted odd ratio for TFs from the Îł event after adjusting for difference in percent identity
of TF genes. Performance of the models was assessed by calculating the r2 between the observed
and predicted odds ratio for each event (ι = 0.87, β = 0.83, γ = 0.65)

186

where we established linear models to predict the degree of retention with the features in Figure
4.2C, a model was built to describe the relationship between the average features of function
groups and degree of retention for each WGD event separately. Beginning with the full set of 34
features, for each WGD event we determined the subset of features (between 5 and 6 in each
case, see Methods) that maximized the F-statistic of the model (see Supplementary File 4.1).
Our models explained 87%, 83%, and 65% of the variance in degree of retention for the ,  and
events respectively (Table 4.1). Applying the F-test to the maximum F-statistic for each model,
we found that each model performs significantly better at the 5% level in explaining the degrees
of retention of function groups than the null model (i.e. fitting the degree of retention to their
mean, Table 4.1).
Features explaining degrees of retention across function groups and WGD events
To determine the importance of individual features in explaining the differences in degree
of retention among function groups, we determined the change in explained variance caused by
independently removing each feature from the model (Table 4.2). Generally, degree of retention
among function groups correlate with higher evolutionary rates (Ka/Ks) within species (paralog)
but lower rate across species (orthologs to one of five species, Figure 4.2B). Degrees of
retentions also negatively correlated with mean expression level and breadth of expression
(Figure 4.2B). However, features with high correlation did not necessarily have a significant
impact on our model performance. In contrast, the features that were retained in multiple linear
models (due to their ability to maximize the F-statistic) have a greater impact on variance
explained when removed. For example, maximum expression (RNA-seq) and expression mean
(AtGenExpress microarray) were, respectively, positively and negatively with duplicate retention
for all three WGD events. This would suggest that functional groups with genes that have more

187

Table 4.1 Performance of best fitting models of the odds ratio of duplicate retention
WGD Event

# Features1

CoD2

F-statistic3

p-value4

Îą

6

0.87

13.8

5.6E-05

β

5

0.83

13.2

7.1E-05

Îł

5

0.65

5.1

7.2E-03

1. The number of explanatory variables (features) used in the best fitting model
2. Coefficient of Determination (r2)
3. The F-statistic is a measure of the goodness of fit of the model to the observed odds ratio.
4. The p-value of goodness of fit based on the F-statistic. A significant p-value (< 0.05)
indicates that the model performs better than fitting the mean value to the data, after
accounting for the number of features in the model.

188

Table 4.2 Importance of features used in the linear models of duplicate retention
Sign1

2

2

2

Expression Mean (AtGenExpress)

-

-0.29

-0.09

-0.49

Expression Maximum (RNASeq)

+

-0.56

-0.59

-0.14

Number of Domains

-

-0.06

-0.36

n/a

Nucleotide Diversity (Pi)

-

-0.06

n/a

-0.32

Expression Correlation (AtGenExpress)

-

n/a

-0.24

-0.21

Expression MAD/Median

-

-0.09

n/a

n/a

Protein Length (in Amino Acids)

+

-0.07

n/a

n/a

Paralog dn

+

n/a

-0.07

n/a

Maximum Percent Identity

+

n/a

n/a

-0.2

Feature

(AtGenExpress)

1. The sign of the association between the feature and duplicate retention
2. Importance of features measured as the decrease in r2 when the feature is removed from
the model, with more negative values indicating greater impact and therefore greater
importance. An ‘n/a’ indicates the feature was not used in the model for that event.

189

specific expression patterns (i.e. lower average across all conditions, but higher maximum
expression under a few specific conditions) tend to retain more duplicates pairs following a
WGD event. Nonetheless, there are a number of cases that defy generalization due to differences
across events. For example, lower expression correlation within function groups was a
significant feature only in the  and  models, while having fewer conserved domains and lower
nucleotide diversity were more important to the  and  models respectively (Table 4.2). These
features more strongly correlated with retention of older duplicate genes suggests long term
retention of duplicates favors genes experiencing stronger purifying selection (low nucleotide
diversity) and those diverged expression patterns (lower expression correlation). The remaining
feature were found in only one of the models and had significant but much smaller impacts the
variance explained (Table 4.2).
Although the degree of retention predicted by the models closely align with the actual
values for each function groups across each event (r2,  = 0.87,  = 0.83,  = 0.65; see Figure
4.2C), our model based on these features alone is obviously imperfect. In particular, degree of
retention for TF were consistently underestimated (black arrows, Figure 4.2C; Supplementary
Figure 4.2), particularly in the  model where the TF odd ratio is predicted to be only 76% of the
actual value. The difficulty of predicting the degree of TF duplicate retention is likely due to the
atypical feature distributions among TFs. For example, the percent base identity between a gene
and its top BLAST hit within A. thaliana is an important predictor of the degree of retention of
duplicates from the  event. This is because the similarity of WGD-duplicates to their best
BLAST for TFs (71.3% identity) resemebles the genome-wide average (72.5%), but the
similarity WGD-singletons to their best hits are significantly higher (Welch’s t-test, pv =
1.9e223) for TF-singletons (66.9%) compared to the genome-wide singleton average (61.3%).
190

This indicates that TFs, once becoming singletons, have higher degrees of sequence conservation
relative to their closest paralog, presumably due to stronger selective pressure, compared to most
other genes. If we inflate the mean difference in perfect identity between WGD-duplicates and
WGD-singleton for TFs by a factor of 2.55 to adjust the decreased difference in percent identity,
the predicted degree of TF retention of the  event becomes 2.94 (red dot, Figure 4.2C),
reducing the error by almost half. In addition to the linear models for predicting degrees of
retention at the function group level, we have established machine learning models incorporating
the same features to predict whether a gene likely has retained duplicate or not. Although this
machine learning model could accurately identify genes with and without retained duplicates, the
overall performance of the model and importance of features varied between kinases, TF and the
rest of the genome. This suggests that, on a gene by gene level, there is dependence between
gene features and functions and, therefore, these models are not useful for explain differences in
the degree of retention between function groups (Supplemental File 4.2).
Taken together, we demonstrated that degree of retention for genes in different function
groups are related to multiple features that are impacted by the timing of WGD events. However,
while these features are useful for predicting the degree of retention for some function groups,
they systematically underestimated degree of retention for TFs. The behavior of TFs departs
from the norm in part because underlying differences in the features of TFs and genome average,
suggesting their evolution following duplication likely differ significantly from other genes.
Partitioning of ancestral expression states following TF duplication
While the gene features (Table 4.2) were generally useful predictors of WGD-duplicates,
they were less useful for predicting TF duplicates specifically. To further explore what
characteristics retained TF WGD-duplicates possess, we examined how the functions of retained

191

TF WGD-duplicates have evolved following WGD events. Approaches to infer ancestral
functions based on those of extant genes have been used to determine the rate of gene activation
and repression in duplicate genes in Drosophila melanogaster (Oakley et al. 2006) and analyze
the evolution of stress response in A. thaliana (Zou et al. 2009b). This approach allows for the
explicit characterization of how duplicate TFs may have deviated (or not) from their ancestral
state over the course of evolution, which in turn may provide information about the
mechanism(s) contributing to TF retention. We first used expression patterns as a proxy of TF
function(s) and inferred the likely expression states of the ancestral TFs prior to WGD (see
Methods). Ancestral expression values were inferred from extant gene expression values that had
been discretized into quartiles (expression state = 0, 1, 2, or 3), based on the distribution of
expression levels for each expression experiment. Additionally, expression data were grouped
into four subsets and analyzed separately, including light and development sets (LightDev),
control conditions (Ctrl), abiotic and biotic stress treatments (Stress), and differential expression
between stress treatments and controls (Diff). This grouping was then used to distinguish
between trends that were universal or specific to certain datasets. We were able to infer 165,385
ancestral expression states across 474 TF WGD-duplicate pairs (a detailed breakdown of inferred
states can be found in Supplementary Table 4.2).
To test how often the ancestral expression states of TFs are retained post-duplication, we
compared the expression states of individual, extant TF WGD-duplicate to its inferred ancestral
states (Figure 4.3A). Although all possible changes in expression state were observed between
ancestral and extant TFs in each expression data subset, the most common ancestral-extant
expression state combination was that the ancestral and extant TFs had the same expression
quartiles (diagonal red boxes, Figure 4.3B). This is true across all expression quartiles, though

192

Figure 4.3 Evolution of expression in TF WGD-duplicates. (A) An illustration of how the zscores in (B) are calculated. Individual TF duplicates are assigned to a cell using the extant (xaxis) and ancestral (y-axis) expression quartile values (dark green = 4th, green = 3rd , yellow =
2nd, white = 1st). Z-scores are then determined by comparing the frequency of the observed
values to frequency distribution that would be expected if expression values were chosen
randomly from a pool of extant and ancestral values. (B) Difference in expression quartile of
individual TFs compared to their ancestors. Heatmaps show the z-scores of the observed
frequency of each difference compared to the expected frequency for LightDev (left column) and
Diff (right column) dataset in three WGD events (ι = top, β = middle, γ = bottom). Darker red
indicates counts further above random expectation. Darker blue indicates counts further below

193

Figure 4.3 (cont’d)
random expectation. (C) An illustration of how the z-scores in (D) were calculated. For each pair
of WGD TF duplicates, the difference in the expression quartile values (colored the same as in
(A)) of an extant duplicate and its ancestral gene is defined as "deviation". Duplicate 1 is the
copy with a higher or equal expression quartile value compared to the other copy (duplicate 2).
The deviation values from each duplicate copy are then used to assign the pair to a cell, where
the duplicate 1 and 2 deviation values are along the x- and the y-axis, respectively. Z-scores are
then determined as in (A). (D) Deviation values of pairs of TF WGD-duplicates from their
ancestral state. Heatmaps show the z-scores of the observed frequency of WGD-duplicate pair
deviation compared to the expected frequency for LightDev (left column) and Diff (right
column) datasets in three WGD events (ι = top, β = middle, γ = bottom). Color correlates with
the magnitude of the z-score as in (A)

194

the deviation from expectation was greatest for expression values in the lowest (1) and highest
(4) quartiles. This general pattern holds across all four data subsets (Supplemental Figure 4.3),
suggesting that most TF WGD-duplicates retain their original expression irrespective of the
expression context. However, when considering a pair of duplicates (Figure 4.3C), we found
that, when the ancestral state was retained in one duplicate, it was lost more often in the other
duplicate than expected by random chance (Figure 4.3D). This may seem to contradict the
results from Figure 4.3B, but we should emphasize that the cases where both duplicates have the
ancestral expression states are still more common (e.g. account for 53% of cases from the ÎąLightDev data set). However, under random permutation of duplicate pairs, 58% of Îą-duplicates
in the LightDev data set are expected to be ancestral-ancestral (Supplemental Table 4.3). In
contrast, we only expected 37% of pairs to be partitioned, but observed 45% pairs to have on
ancestral and one non-ancestral expression state. We find the same trend for duplicates from
Control, Stress, and Diff data sets originating from the relatively younger ι and β events (see
Supplemental Table 4.3).
Influence of the timing of TF duplication and expression state evolution
The “partitioned” state of TF WGD-duplicates pairs is over-represented at higher degrees
for more recent WGD events (Figure 4.3D). In the relatively older WGD events (β and γ),
having neither duplicate inherit the ancestral expression state is more common than the
partitioned state where only one copy inherits the ancestral state. Using ANOVA, we confirmed
that there is indeed significant interaction between the expression state of a TF WGD-duplicate
pair and the timing of the WGD event (pv < 2e-16), which indicated that partitioning occurred
relatively quickly after the most recent WGD, but that these partitioned patterns were not
necessarily retained as the duplicates age.

195

Next we asked if TF duplicate expression levels tend to increase or decrease when they
deviate away from the ancestral state. Because we found a significant interaction between the
expression state evolution of TF WGD-duplicate pairs and the subset of the expression data used
(pv < 3e-05), we asked this question using each expression data subset. For the LightDev
(Figure 4.3D), Ctrl, and Stress expression subsets (Supplemental Figure 4.4), partitioning of
ancestral expression states among duplicates favors small, negative changes from the ancestral
states. Based on an earlier study showing that A. thaliana up-regulation in response to stress is
lost more frequently that down regulation (Zou et al. 2009b), we anticipated TFs would most
often lower expression quartiles compared to their ancestral state. However, when we looked at
the Diff subset (the contrast between samples in the Stress subset and their respective controls)
we found that TFs were equally likely to increase or decrease differential expression in response
to stress compared to the ancestral state, in contrast to absolute expression levels where decrease
is the norm.
To further assess how what rates of evolution are from ancestral expression states to
extant states, we modeled the transition from ancestral expression (O) to higher (+) and lower (-)
expression states following a WGD duplication event (see Methods). We compared a oneparameter model where the rates of transition from (O) to (+) and (-) were equal to a twoparameter model where the rates from (O) to (+) and (-) were allowed to differ (Supplemental
Figure 4.5). The two-parameter model was significantly better than the one parameter model
when absolute expression levels are considered using the LightDev (likelihood ratio test, pv =
2.2e-11), Ctrl (pv = 2.7e-3), and Stress (pv = 2.9e-3) subsets. For these subsets, the rate of
evolution from (O) to (-) was 1.9~3.1 times more frequent than that from (O) to (+). For the Diff
subset, (O) to (-) was only 1.2 times more frequent, which was not significant (pv = 0.43). In

196

summary, these results suggest that the evolution of TF duplicates favors decreasing expression
levels relative to the ancestral expression state (Control, LightDev, and Stress). However, when
looking at differential expression in response to stress, TF duplicates can evolve in either
direction with approximately equal likelihood. Thus, following duplication, TF duplicates may
have increased or decreased responses to stress, rather than losing the response altogether.
Asymmetry in the partitioning of ancestral expression and regulatory sites
Thus far we show that an ancestral expression state tends to be retained by only one copy
of a TF WGD-duplicate pair and each expression state is considered individually. One
outstanding question is whether each copy would retain different parts of the ancestral
expression state, as would be expected if the TF duplicates were retained due to
subfunctionalization (Force et al. 1999). To address this, we considered all the partitioned
expression states (i.e. all expression series showing partitioning) across a pair of TF WGDduplicates. If partitioning were random, the number of ancestral states retained by a single
WGD-duplicate is expected to follow a binomial distribution for the given number of partitioned
expression states and a retention probability of 0.5 (both copy equally likely to retain ancestral
states). Under this scenario, the expected asymmetry of a duplicate pair (the difference in the
fraction of ancestral states inherited between duplicates) is 0.18 (Figure 4.4A). However, the
actual mean asymmetry between TF WGD-duplicates was 0.68, significantly different from
random partitioning (pv < 1e-323) expected under the subfunctionalization model (Figure 4.4B).
As with mean asymmetry, the skewed distribution of asymmetry values is also significantly
different from what was expected from random partitioning (Kolmogorov–Smirnov test, pv <
2.2e-16). This biased partitioning was also found within the Ctrl (mean asymmetry= 0.84),
LightDev (mean = 0.67), Stress (mean = 0.70), and Diff (mean = 0.56) subsets. To assess the

197

Figure 4.4 Asymmetry of ancestral state retention in TF WGD-duplicates. (A) Example of
how Asymmetry score (Asy, see Methods) is calculated. Ancestral conditions are indicated by
yellow boxes and non-ancestral conditions by grey boxes. Among a pair of duplicates, an
‘ancestral’ copy (red arrow) is the duplicate retains more ancestral states than the other, ‘nonancestral’ copy (blue arrow). In case where equal numbers of ancestral states are inherited (the

198

Figure 4.4 (cont’d)
first case with Asy=0), the ancestral and non-ancestral designation is assigned randomly. (B) The
Asymmetry scores of ancestral expression partitioning between TF WGD-duplicates. Red
columns indicate the expected frequency of each score bin based on a series of grouped
Bernoulli trials (see Methods) while blue columns indicated the observed frequency. (C) The
Asymmetry scores of ancestral cis-regulatory site partitioning between TF WGD-duplicates. Red
and blue columns are as described in (B). (D) The frequency distribution of the difference in
number of novel cisÂŹ-regulatory sites between ancestral and non-ancestral WGD duplicate
copies. The value on the x-axis is calculated as the number of novel regulatory sites in the nonancestral copy minus the number in the ancestral copy.

199

possibility that the observed pattern of partitioning may result from non-independent loss of
ancestral expression due to the use of correlated time course data, we assembled subsets of
LightDev, Stress, and Diff conditions by using only the first or last time point in each time
course and found that the asymmetry scores for these subsets were virtually unchanged from
those using the full datasets, the first time points (LightDev = 0.68, Stress = 0.73, Diff = 0.58) or
the last (LightDev = 0.68, Stress = 0.71, Diff = 0.59) time points. Given these results, for each
TF WGD-duplicate pair, we can generally define one duplicate as being “ancestral” and the other
as being “non-ancestral”.
Why then is the non-ancestral copy being retained? One hypothesis is that the nonancestral copy is retained because it has acquired a novel function in the form or new expression
or regulation. To test this, we first applied our model of ancestral-state partitioning to cisregulatory sites. We used cis-regulatory sites here because the discretized expression levels used
above allowed us to determine the direction of changes away from the ancestral expression state,
but not whether an expression state was novel. The cis-regulatory sites used here are from
putative binding sites of 345 A. thaliana TFs (O’Malley et al. 2016). We applied the same
methodology used to infer ancestral gene expression to infer ancestral cis-regulatory sites of
ancestral TFs (see Methods). In 16,015 cases, we found a cis-regulatory site in either one of the
duplicate copies or the ancestral genes. Of these, in 57.8% of cases, an ancestral site was lost in
one duplicate, while 10.5% and 16.2% cases we saw the ancestral cis-regulatory site lost or kept
in both duplicates, respectively Similar to what we see for the partitioning of an ancestral
expression state (Figure 4.3), loss of an ancestral cis-regulatory site in only one copy occurs
more often than what would be expected if WGD-duplicate and ancestral genes were randomly
associated (42.3%; t-test, pv < 1e-323). In contrast, retention (expected = 24.0%, pv < 1e-323)

200

and loss (expected = 18.5%, pv < 1e-323) of ancestral cis-regulatory sites in both WGDduplicates were significantly less frequent than randomly expected. In addition, similar to
ancestral expression state evolution, the partitioning patterns of ancestral cis-regulatory sites
were highly asymmetric (Figure 4.4C), significantly different from random partitioning
(Kolmogorov–Smirnov test, pv < 2.2e-16). Thus, much like what we observed for expression, TF
WGD-duplicates can be classified into ancestral and non-ancestral copies with regard to cisregulatory sites. Most importantly, amongst the 249 duplicate pairs with at least one novel
regulatory site, in 71.0% of cases the non-ancestral copy had more novel cis-regulatory sites
(Figure 4.4C), significantly higher than random expectation (49.8%, pv < 3.8e-12). Furthermore,
in 61.8% of duplicate pairs the novel cis-regulatory sites are only found in the non-ancestral
copies, compared to 14% of pairs where all of the novel sites are in the ancestral copies. These
patterns suggested that, the acquisition of novel cis-regulatory sites likely contribute to the
retention of the non-ancestral TF duplicate copies. This conclusion may be similar if we consider
novel expression states, considering that the ancestral and non-ancestral designation defined
according to expression data tend to have the same designation based on cis-regulatory sites
(59.8%, compared to expected by random association at 24.6%, pv = 1.8e-20).
Within this pool of duplicate pairs with distinct ancestral and non-ancestral duplicates,
there are a number of examples the non-ancestral copies exhibiting a different function from the
ancestral duplicate. The non-ancestral gene KNAT4 has gained 37 novel regulatory sites relative
to KNAT3, which has retained all partitioned expression and regulatory sites (Figure 4.5A). In
this case, both genes retain a development regulatory function, but KNAT4 is primarily found in
the elongation zone of roots in phloem and pericycle cells, while KNAT3 is found in the
differentiation zone in pericycle and cortex cells (Truernit et al., 2006, Truernit and Haseloff,

201

Figure 4.5 Expression partitioning between duplicate pairs with high regulatory
asymmetry. Expression partitioning of three duplicate pairs KNAT3/4 (A), BCP2/3 (B),
DAG1/2 (C) where the non-ancestral duplicate (blue arrow) exhibits differential function from
202

Figure 4.5 (cont’d)
the ancestral duplicate (red arrow). Expression quartile is indicated by color (dark green = 4th,
green = 3rd, yellow = 2nd, white = 1st). Note that only expression conditions under which
function differs between the duplicates are shown

203

2007). In another example, BPC3 is a non-ancestral duplicate which has 20 novel regulatory sites
and has lost ancestral expression in 15 conditions where it is retained in its duplicate, BPC2
(Figure 4.5B). Previous research found BPC3 functions antagonistically not only to BPC2, but
other members of BPC regulatory family in regard to controlling growth, leaf shape, and flower
development (Monfared et al., 2011). Finally, the non-ancestral copy DAG2 is a positive
regulator of phyB induced germination which is directly regulated by the ancestral copy DAG1
(Santopolo et al. 2015). This is of particular interest because, in spite of having opposite
regulatory roles, both duplicates have similar expression breadth (Gualbertia et al 2002) and our
own data indicates that ~40% of inferred ancestral cis-regulatory elements are conserved in both
copies even though the DAG1 retains most of the ancestral response to light (Figure 4.5C). This
indicates that function differentiation can arise even when ancestral expression is incompletely
partitioned between copies.
Patterns of WGD-duplicate divergences and partitioning results from evolutionary bias
We have demonstrated that partitioning of ancestral expression and regulation into
ancestral and non-ancestral duplicates is favored following duplication of TFs. It remains an
open question if this ancestral state partitioning is maintained and thus the duplicate retains the
ancestral expression/regulation is likely under selection. Alternatively, if the rate of ancestral
state loss of the second copy is similar to that of the first, it would suggest the partitioning is
simply a transition state and is not maintained. To determine which of the above cases is likely
true, we modeled loss of ancestral states of TF WGD-duplicate pairs (see Methods). Using the
synonymous substitution rate (ds) of TF WGD-duplicate pairs derived from the , , or  events
as a proxy for time, the rate of transition between WGD-duplicate pairs where neither (state O),
only one (state I), or both (state I) duplicates had lost ancestral expression was modeled (Figure

204

4.6A). We compared a model where the rates of transitions between all states were equivalent
(same rates for losing the ancestral states in both duplicates, one-parameter model) with a model
where the transition rates between state O and I were allowed to vary from those between state I
and O (two-parameter model). These models were applied to all expression subsets
(Supplemental Figure 4.6), the results of the one and the two parameter models using the
LightDev dataset are shown as an example in Figure 4.6B.
We found the two-parameter model to be significantly better at explaining the observed
difference in WGD-duplicate states over time (Likelihood Ratio Test, p-value < 2e-14).
Regardless of the expression data set, the transition rates between state O (ancestral expression in
both duplicates) and I (ancestral expression in on duplicate) were 7-13 times higher than the rates
between state I and II (ancestral expression in neither duplicate) (Figure 4.6). In addition, based
on the estimated rates of ancestral state loss over time by the two-parameter model (curves in
Figure 4.6B), the number of partitioned WGD-duplicates accumulated rapidly post WGD,
followed by a relatively slow accumulation of bases where ancestral expression states had been
lost in both duplicates. We also assessed a four-parameter model (OI, III, III, IO) that
was not better than the two-parameter model. Applying this same approach to model regulatory
site evolution revealed that the best fit model for regulatory site evolution involved allowing all
four rate parameters to vary (p-values 4.8e-13 and 1.2e-11 vs. one and two-parameter models,
respectively; Figure 4.6C). The rates governing the OI transition (x) are two orders of
magnitude higher than the III transition (w, Figure 4.6D). Importantly, in the four-parameter
model, there was a high rate of OI transition estimated at the early stage of WGD (blue curve,
Figure 4.6C). In addition, an appreciable proportion of partitioned duplicates lost ancestral
regulatory sites in the second copy (green curve, Figure 4.6C) that contributed to the pattern

205

Figure 4.6 ODE models of TF WGD-duplicate expression and cis-regulatory site evolution
relative to the ancestral state. (A) In this model, we consider the transition of WGD-duplicate
pair expression states between three possible scenarios (O = both retained, I = one retained, II =
neither retained) using four variables representing the rate of transition between state (x,y,w,z).
(B) Left and middle: results for the one parameter (x=y=w=z) and two parameter (x=y|w=z)
versions of the expression state model showing the change in time (x-axis) and the frequency (yaxis) of each scenarios. Curves represent the continuous output of the model in different
scenarios. The significance of including additional parameters (the p-value between the curly
brackets) was determined using the likelihood ratio test. Right: A bar graph of the parameter
values for the one (orange) and two (green) parameter versions of the expression ODE model.
(C) Left three sub-graphs: results for the one parameter (x=y=w=z), two parameter (x=y|w=z),
and four parameter (x|y|w|z) versions of the cis-regulatory site model showing the change in time

206

Figure 4.6 (cont’d)
(x-axis) and the frequency (y-axis) of each WGD-duplicate-pair scenario. Curves represent the
continuous output of the model in different scenarios. The p-values are derived from the
likelihood ratio tests between models. Far right: a bar graph of the parameter values (x,y,w,z) for
the one (orange), two (green), and four (blue) parameter versions for the cis-regulatory site ODE
model.

207

where the proportion of partitioned duplicated peaked at low ds followed by a reduction. This is
in sharp contrast compared to the transition rate estimate over time for expression where second
copies tend not to lose ancestral expression state (Figure 4.6B), indicating that regulatory sites
are faster evolving and more labile compared to expression states.

208

CONCLUSIONS
In this study, we have shown that duplicates are retained at different rates across function
groups. In addition, we established linear models to assess how expression, conservation, and
sequence structural features of genes in these functional groups may explain their retention rate
difference. Although the linear model is far from perfect, it serves as the basis for exploring more
complicated interactions underlying duplicate retention, i.e., the potential interaction between
gene features and annotated gene function suggested by our results. We also demonstrate a
preference for maintaining partitioned expression and regulatory site states between TF WGDduplicate pairs. Yet, while we have established that retained duplicate genes have distinct
expression, sequence and regulatory features and TF duplicate genes in particular are
characterized by asymmetric-partitioning, the question of what this implies about why duplicate
genes are retained remains to be addressed.
Many mechanisms have been proposed to explain why duplicate genes are retained. Any
duplicate pair could potentially be retained via neofunctionalization (Ohno,1970)) or escape from
adaptive conflict (Des Marais and Rausher 2008) which involve the evolution of new or
improved function that is positively selected for. However, subfunctionalization (Force et al.
1999) or gene balance (Birchler and Veitia 2007; Birchler and Veitia 2010; Baker et al. 2013) are
specific to TFs and other gene with a large number of interactions/functions (Seoighe and
Gehring 2004; Maere et al. 2005; Shiu et al. 2005, Alvarez-Ponce and Fares 2012) which all
need to maintained following WGD. On an experiment-by-experiment basis, the partitioning of
ancestral expression states (Figure 4.3D) would appear to support the notion of WGD-duplicate
retention by subfunctionalization (Force et al., 1999). However, when examining the ancestral
state partitioning patterns across multiple experiments, we find an extreme bias where one TF

209

duplicate retains most of the ancestral states and the other, non-ancestral copy retains few or
none (Figure 4.5). Most importantly, we showed that the non-ancestral copy tends to gain novel
cis-regulatory sites (Figure 4.5D) and exhibit differential expression from the ancestral state.
This pattern harkens back to the notion of there being an ancestral copy and a neofunctionalized
copy after duplication, contributing to the retention of both duplicates (Ohno, 1970). This would
appear to be the case for duplicate pairs like KNAT3/4, BPC2/3, and DAG1/2.
Nonetheless, we should note that there remain asymmetrically partitioned duplicates that
are retained without clear evidence of neofunctionalization. A clear case of this is the duplicate
pair ANAC19/72 which, in spite of ANAC72 gaining novel 21 regulatory sites, appears to have
redundant function regulating stress response, both with each other and with others NAC-family
TFs (Tran et al., 2004; Zheng et al., 2012; Takasaki et al., 2015). It has been theorized that
seemingly redundant duplicates may be retained due to subfunctionalization at the expression
level following a reduction expression after duplication and/or subsequent “rebalancing” of
expression that could be positively selected (Qian et al., 2011). Yet while this might explain the
retention of asymmetric duplicates with similar function, it cannot explain the retention duplicate
pairs where ancestral expression in maintained across both copies. For example, 7.5% and 13.9%
of duplicates TF pairs have retained >80% of ancestral expression in both copies in the Stress
and LightDev data set respectively. Thus, while neofunctionalization and subfunctionalization
may explain the retention of partitioned duplicates, the presence of duplicate pairs with a high
degree of ancestral expression in both copies and the overall prevalence of retaining ancestral
expression following the ι and β WGD events (Supplemental Table 4.3) remains to be
addressed.

210

If subfunctionalization does not play a predominant role in TF duplicate retention, what
other mechanisms may be responsible? One prominent hypothesis is gene balance where
stipulate that duplicate genes with products that form multimeric complexes will tend to be
retained to maintain the stoichiometry (Birchler and Veitia 2007; Birchler and Veitia 2010) and
enables future sub- and/or neofunctionalization (Veitia et al. 2013). If gene balance does play a
significant role in retention of TF duplicates, we would expect duplicates from more recent
WGD events to have higher proportion of cases where both copies retained the ancestral
expression and regulatory site states. However, our ODE model for evolution of duplicate TF
pairs indicate that the proportion of duplicates both retaining ancestral states reduces quickly
following WGD (Figure 4.6), suggesting that, if gene balance plays a significant role, it is
limited to the initial period after WGD. The caveat is that our current ODE models of ancestral
expression and regulatory site evolution is based on WGD events that are >50 million old. It will
be useful to incorporate data from other species with more recent WGD events into the model to
further address this question. Additionally, WGD-Duplicates TFs are known to be preferentially
retained across many plant species (Carretero-Paulet and Fares 2012), it will be of interest to see
if the patterns of ancestral expression and regulatory site partitioning we have uncovered in A.
thaliana are shared in other plant lineages sharing the , , and  events, or common to any
species with ancient WGD events. Furthermore, our study focuses on the overall pattern of TF
evolution. It is anticipated that different TF families will evolve differently from each other. In
future studies, it will be important to directly compare the size, rate of retention, and rate of
partitioning both within and across species in individual families.

211

MATERIALS AND METHODS

Genome sequences, gene annotation, and Expression Data
Genome sequences, protein sequences, and gene annotation information for A. thaliana
was obtained from Phytozome v10 (https://phytozome.jgi.doe.gov/pz/portal.html). WGDs were
defined according to Bowers et al. (2003) and tandem genes in A. thaliana were defined as pairs
of reciprocal best BLAST hits with an e-value < 1e-10 and ≤ 5 intervening genes. Expression
microarray data for this study was taken from AtGenExpress (Schmid et al. 2005; Kilian et al.
2007; Goda et al. 2008), normalized using RMA (Irizarry et al. 2003) in R as performed
previously (Zou et al. 2009a). The array data was divided into four groups: control conditions (in
environmental condition experiments, Ctrl), light and development set (LightDev), abiotic and
biotic stress treatments (Stress), and differential expression between stress treatments and
controls (Diff) (Supplemental Table 4.4). The Diff data contains the log2 normalized difference
between data sets for each stress condition/treatment/duration and its corresponding controls. In
addition to microarray data, we have included a set of 214 RNA-sequencing samples
(Supplemental Table 4.5) from A. thaliana Col1 wildtype from the Sequence Read Archive
(https://www.ncbi.nlm.nih.gov/sra) as of September 30, 2014. Raw sequence reads were
processed using Trimmomatic (Bolger et al. 2014), with a quality threshold of 20, window size
of 4, and hard-clipping length of 3 for leading and trailing bases. Processed reads were then
mapped to the A. thaliana genome using Tophat2 (Kim et al. 2013) and expression levels
calculated with Cufflinks (Trapnell et al. 2010), both with a maximum intron length of 5,000bp.

212

Defining TFs and other groups of genes in A. thaliana
TFs were defined according to the criteria used by the Plant Transcription Factor
Database (Jin et al. 2014) with 1,717 annotated TF loci in A. thaliana. To assess the degrees of
TF duplicate retention after each WGD event, we defined a set of “functional groups” for
comparison following from the procedure used in Maere et al. (2005). To compare among genes
with divergent functions and to ensure the log odds indicative of the degrees of retention could
be defined for each group, function groups were defined using Gene Ontology (GO) (Ashburner
et al. 2000) terms in the molecular function and biological process categories from The
Arabidopsis Information Resource (https://www.arabidopsis.org/), and only groups containing
100-2,000 genes and ≥20 WGD-duplicate pairs were kept. We excluded GO:0006355 (regulation
of transcription, DNA-templated) due to its substantial overlap with the TF group we have
defined above. The remaining 19 function groups include: ATP Binding (GO:0005524), catalytic
activity (GO:0003824), defense response (GO:0006952), DNA endoreduplication
(GO:0042023), hydrolase activity hydrolyzing O-glycosyl compounds (GO:0004553), kinase
activity (GO:0016301), lipid binding (GO:0008289), oxidoreductase activity (GO:001649),
oxygen binding (GO:0019825), protein binding (GO:0005515), proteolysis (GO:0006508),
response to auxin (GO:0009733), response to chitin (GO:0010200), RNA binding
(GO:0003723), transferase activity, transferring glycosyl groups (GO:0016757), translation
(GO:0006412), transporter activity (GO:0005215), ubiquitin-protein transferase activity
(GO:0004842), zinc ion binding (GO:0008270). A list of genes in each group can be found in
Supplemental Table 4.6.

213

Fitting odds ratio of duplicate retention within each group of genes for each WGD event
using linear models
A gene was designated as a "WGD-duplicate" if its paralog derived from a particular
WGD event is present. For a gene without its paralog from WGD, it was designated as a "WGDsingleton" gene. The degree of retention for a function group, g, after a specific WGD event, w,
is defined as:
𝑅𝑔,𝑤 =

(𝐷𝑔,𝑤 /𝑆𝑔,𝑤 )
(𝐷¬𝑔,𝑤 /𝑆¬𝑔,𝑤 )

Where Dg,w and DÂŹg,w are the numbers of WGD-duplicate genes in group g and those not in
group g (ÂŹg), respectively. Sg,w and SÂŹg,w are the numbers of WGD-singleton genes in group g
and those not in group g (¬g), respectively. The 95% confidence interval around the pointestimate Rg,w was defined using the “fisher.exact” function in R, the details of which can be
found at in Fay (2010). For each WGD event, we established a general linear model with the glm
function in the R environment which relates the Rg,w to a set of features of each gene group. The
34 features (predictor variables, Supplemental Table 4.1) were filtered with the following
procedures to prevent over-fitting because we have only 20 function groups. We calculated the
correlation between all features to find all cases where the absolute value of correlation was
>0.7. The considerations for which features to keep included: (1) how well each feature
correlated with Rg,w on its own, (2) whether the feature was derived from a subset of another
feature, and (3) the number of other features with a correlation > 0.7 (favored the elimination of
more features). In addition to the above criteria, one data set (protein-protein interactions) was
eliminated because of a high frequency of missing values (88%). The synonymous substitution
rate (KS) feature and any feature using KS in their calculation were also excluded because they
would be highly correlated with WGD timing and confound our analyses comparing the three
214

WGD events. The filtering step left 11 features for building the general linear model. Following
fitting the glm function, features were ranked according to their p values from the least to the
greatest and the feature with the largest p-value was dropped. The model was then fit to the
reduced feature set and features were once again ranked. This process was repeated until the Fstatistic (a measure of goodness of fit of the given model against a null model where all
coefficients are set to zero) of the model was maximized and the final p value was calculated
based on the maximal F-statistic. The final model for each event can be found in Supplementary
File 4.1.
Inferring ancestral expression levels and cis-regulatory sites
DNA-binding domains were identified in TF protein coding sequences using hmmscan
via HMMER3 (Mistry et al. 2013) based on the Pfam-A version 29.0 HMMs (Finn et al. 2016)
with a threshold e-value of 1e-5. TFs were classified into families according to their DNAbinding domains and 44 of 59 TF families with ≥4 members were used for further analysis
(Supplemental Table 4.7). For each TF family, full-length protein sequences were aligned using
MAFFT (Katoh and Standley 2013) with default parameters. The phylogeny of each TF family
was obtained using RAxML (Stamatakis 2014) with the following approach: rapid Bootstrapping
algorithm, 100 runs, GAMMA rate heterogeneity, and the JTT amino-acid substitution model.
These trees were then mid-point rooted with retree in PHYLIP (Felsenstein, 1989) and used to
infer the ancestral gene expression states and the cis-regulatory sites of WGD-duplicate TF pairs
with BayesTrait (Pagel et al. 2004) as was done in our earlier study (Zou et al. 2009a). The
expression data sets used are described in Supplemental Table 4.4. The discretized gene
expression state (0,1,2,3) was based on the quartiles of gene expression levels within each
experiment. Thus the inferred, ancestral expression state was also discretized. For cis-regulatory

215

sites, the binding targets of 345 A. thaliana TFs were defined based DNA Affinity PurificationSequencing data (O’Malley et al., 2016) from the Plant Cistrome Database
(http://neomorph.salk.edu/dap_web/pages/index.php) where at least 5% of the read associated
with a site were found to be in the 200bp peak region. We inferred whether a particular site was
present or absent (0,1) in the common ancestor of a duplicate pair. For both expression and
regulatory site data, in cases where there was a missing value, it was explicitly included as an
ambiguous state. To call the ancestral state from the expression or cis-regulatory site data, we
required a posterior probability > 0.5. Cases where the called state was ambiguous or no majority
existed were excluded from further analysis.
Asymmetry of the retention of ancestral expression and regulatory sites
For determining expression state asymmetry, only TF WGD-duplicates with ≥5
partitioned ancestral expression states in one of the four expression datasets (Ctrl, LightDev,
Stress, and Diff) were considered. For a WGD-duplicate pair with genes A and B, if the number
of inherited ancestral expression states in A was larger or equal to that in B, then A and B were
defined as the ancestral and the non-ancestral duplicate copies, respectively. The degree of
asymmetry (YA,B) of expression states between two duplicates was defined as:
𝑌𝐴,𝐵 = max(𝐹𝐴 , 𝐹𝐵 ) − (1 − max(𝐹𝐴 , 𝐹𝐵 ))
Where FA and FB are the frequency with which ancestral expression was retained for duplicates
A and B, respectively. By definition, FA + FB = 1, such that YA,B has value between 0 (when FA =
FB, no asymmetry) and 1 (when either FA or FB = 1, maximum asymmetry).
With the asymmetry values for each TF pair, an average asymmetry value of all TF pairs
was calculated for each expression dataset, as well as for the union of all TF duplicates from all
datasets (1,239 values total) to assess how the observed degree of asymmetry compared to what

216

would be expected from if every partitioned state was independent (i.e. each gene has an equal
chance of retaining the ancestral state regardless of the outcome of previous partitioning events).
We also defined two subsets of the LightDev, Stress, and Diff data sets using the first and last
element of each times series respectively because the expression of genes at different points of a
time series are potentially correlated. The number of genes with >5 partitioned conditions genes
decreased in the subsets of LightDev (all = 334, first = 327, last = 325), Stress (all = 347, first =
265, last = 272), and Diff (all = 351, first = 277, last = 269) data sets. We excluded the Ctrl data
set because it is composed of only four series, mean that no genes could pass the >5 partitioned
condition cutoff.
The expected distribution of asymmetry values for the expression states of TF WGDduplicates (under the assumption of independent of partitioning events) was determined by
conducting a series of Bernoulli trails equal to the total number of partitioned states amongst TFWGD duplicates. In each of these trials there was an equal probability that either the first or
second duplicate receive the ancestral state. The results of these trials were then grouped
according the exact per gene distribution of partitioned states in TF-WGD duplicates and an
asymmetry value was calculated for each group. This procedure was repeated 1,000 times using
an independent set of trials and subsequent groupings
For assessing cis-regulatory site asymmetry, only TF WGD-duplicates with ≥5 inferred
ancestral cis-regulatory sites we considered (402 WGD-duplicate pairs total). Similar to
expression state asymmetry, in each duplicate pair the ancestral and non-ancestral duplicates
were defined according to the number of inherited ancestral sites. For each WGD-duplicate pair,
the degree of asymmetry of cis-regulatory site among a TF pair was defined analogous to what

217

was done for expression. The expected distribution of asymmetry values for the cis-regulatory
sites of TF WGD-duplicates was determined using the same procedure as for expression states.
Ordinary differential equation models of TF state evolution
The change in expression states from the ancestral expression quartile to either a higher
or lower quartile in an extant TF was modeled as a system of ordinary differential equations such
that:
−(𝑥 + 𝑦) 𝑤
𝑑 𝑂
𝑥
−𝑤
(+ ) = (
𝑑𝑡
−𝑦
0
−

𝑧
𝑂
0 ) (+ )
−𝑧 −

Where O, +, and - are the frequency of TF WGD duplicate genes retaining the ancestral
expression states, having a higher-than-ancestral expression level, and having a lower-thanancestral expression level, respectively. The parameters x, y, w, and z, define the transition rates
between these states. This system of equations was solved in Maxima
(http://maxima.sourceforge.net/index.html) and best parameters for the observed distribution of
duplicates pairs were determined using maximum likelihood estimates calculated with the bbmle
package in R (https://cran.r-project.org/web/ packages/bbmle/index.html). Non-linear
minimization was used to approximate an initial guess, although the actual initial parameters
often needed to be adjusted to reach a convergent solution. The best fit parameters for this single
duplicate expression state evolution model can be found in Supplemental Table 4.8.
The loss of ancestral expression states in a pair of duplicated TFs was modeled as a
system of ordinary differential equations such that:
−𝑥
𝑑 𝑂
(𝐼) = ( 𝑥
𝑑𝑡
𝐼𝐼
0

𝑦
−(𝑦 + 𝑤)
𝑤

0
𝑂
𝑧 )(𝐼 )
−𝑧 𝐼𝐼

Where O, I, and II are the frequency of TF WGD duplicate pairs where both, one, or neither
duplicate retained the ancestral expression state. The parameters x, y, w, and z, define the
218

transition rates between these states. This system of equations was solved and the initial and best
parameters were estimated in the same fashion as above. The best fit parameters for this pairwise
expression state evolution model can be found in Supplemental Table 4.9. The same model was
also applied to ancestral regulatory sites with O, I, and II representing the frequency of TF WGD
duplicate pairs where both, one, or neither duplicate retained the ancestral regulatory site.

219

ACKNOWLEDGEMENTS
We thank Johnny Lloyd and Zing Tsung-Yeh Tsai for their advice regarding modeling
duplicate retention and analyzing the importance of predictive features.

220

APPENDIX

221

Supplemental File 4.1: Linear equations used to predict odds of duplicate of retention in
different WGD events across function groups

𝑂𝑑𝑑𝑠(𝛼) =

−0.370 ∗ (𝐸𝑥𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝑀𝑒𝑎𝑛, 𝐴𝑡𝐺𝑒𝑛𝐸𝑥𝑝𝑟𝑒𝑠𝑠)
+ 4.763 ∗ 10−4 ∗ (𝐸𝑥𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝑀𝑎𝑥, 𝑅𝑁𝐴)
−140.6 ∗ (𝑁𝑢𝑐𝑙𝑒𝑜𝑡𝑖𝑑𝑒 𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦, 𝜋)
−1.073 ∗ 10−3 ∗ (𝑃𝑟𝑜𝑡𝑒𝑖𝑛 𝐿𝑒𝑛𝑔𝑡ℎ)
−2.325 ∗ (𝐸𝑥𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝑀𝐴𝐷/𝑀𝑒𝑑𝑖𝑎𝑛, 𝐴𝑡𝐺𝑒𝑛𝐸𝑥𝑝𝑟𝑒𝑠𝑠)
−0.190 ∗ (𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐷𝑜𝑚𝑎𝑖𝑛𝑠)
+4.786

𝑂𝑑𝑑𝑠(𝛽) =

−0.294 ∗ (𝐸𝑥𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝑀𝑒𝑎𝑛, 𝐴𝑡𝐺𝑒𝑛𝐸𝑥𝑝𝑟𝑒𝑠𝑠)
+ 6.745 ∗ 10−4 ∗ (𝐸𝑥𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝑀𝑎𝑥, 𝑅𝑁𝐴)
−4.103 ∗ (𝐸𝑥𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝐶𝑜𝑟𝑟𝑒𝑙𝑎𝑡𝑖𝑜𝑛, 𝐴𝑡𝐺𝑒𝑛𝐸𝑥𝑝𝑟𝑒𝑠𝑠)
+2.686 ∗ (𝑃𝑎𝑟𝑎𝑙𝑜𝑔 𝐷𝑛 )
−0.484 ∗ (𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝐷𝑜𝑚𝑎𝑖𝑛𝑠)
+4.130

𝑂𝑑𝑑𝑠(𝛾) =

−0.806 ∗ (𝐸𝑥𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝑀𝑒𝑎𝑛, 𝐴𝑡𝐺𝑒𝑛𝐸𝑥𝑝𝑟𝑒𝑠𝑠)
+ 4.329 ∗ 10−4 ∗ (𝐸𝑥𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝑀𝑎𝑥, 𝑅𝑁𝐴)
−587.5 ∗ (𝑁𝑢𝑐𝑙𝑒𝑜𝑡𝑖𝑑𝑒 𝐷𝑖𝑣𝑒𝑟𝑠𝑖𝑡𝑦, 𝜋)
−5.553 ∗ (𝐸𝑥𝑝𝑟𝑒𝑠𝑠𝑖𝑜𝑛 𝐶𝑜𝑟𝑟𝑒𝑙𝑎𝑡𝑖𝑜𝑛, 𝐴𝑡𝐺𝑒𝑛𝐸𝑥𝑝𝑟𝑒𝑠𝑠)
−0.133 ∗ (𝑀𝑎𝑥𝑖𝑚𝑢𝑚 𝑃𝑒𝑟𝑐𝑒𝑛𝑡 𝐼𝑑𝑒𝑛𝑡𝑖𝑡𝑦)
+5.282

222

Supplemental File 4.2: Predicting WGD-duplicate retention status of individual genes using
machine learning
Machine learning models for TFs, kinases, and all genes in the genome were generated to
predict whether a gene in a particular group had a retained WGD paralog from either the , ,
and  event, as small numbers of β and γ made the difficult to correctly classify on their own..
The machine learning was performed using the Random Forest algorithm implement in the R
package randomForest (https://cran.r-project.org/web/packages/randomForest/index.html). We
filtered the gene level feature set from a previous study (Lloyd et al. 2015) by removing those
with missing values for ≥5% of genes. For the remaining features, missing values were imputed
with the rfImpute algorithm in randomForest using 10 iterations of 500 trees. The final matrix of
genes and features for TFs, kinases, and the whole genome can be found in Tables S7, S8, and
S9, respectively. Using the imputed data set for each group of genes and for each WGD event,
we ran the Random Forest algorithm 10 times with 500 trees (each time with 10 fold cross
validation) and collected the resulting votes (retained or not) for constructing Receiver Operating
Characteristic curves (ROCs). The importance of each individual feature was assessed using
Mean Decrease in Accuracy (MDA), the average number of genes misclassified across multiple
runs as a result of removing the feature in question. The statistical significance of the difference
in values of a feature between WGD-duplicates and WGD-singletons was determined using
Welch’s t-test.
To evaluate the performance of our classifiers, we determined receiver operating
characteristic curves (ROCs) for each model (Fig 1) and calculated the Area Under Curve (AUCROC), a metric that summarizes the ability of the classifier to recover true positive WGDduplicate genes at different false positive rates. An AUC-ROC of 0.5 indicates that the classifier

223

is no better than randomly labeling genes as having a retained duplicate or not, while an AUCROC of 1.0 indicates that the classifier can make predictions without error. Among the
classifiers, the one characterizing the full genome performed best (AUC-ROC = 0.86), followed
closely by protein kinases (AUC-ROC = 0.82), while the classifier for TFs, while much better
than random, did not perform as well (AUC-ROC = 0.74). To investigate the source of the
difference, we determined the importance of each feature to the classifier by calculating the
Mean Decrease in Accuracy (MDA) which is the average number of genes misclassified across
multiple runs as a result of removing a feature (Supplemental Table 4.10). Given TFs are the
least well predicted, we suspected the informative features for predicting retention in TFs would
differ greatly from those for the genome at large and the protein kinases. Contrary to this
expectation, the ranking of importance for TF WGD-duplicate prediction was more similar to the
ranking of features for the whole genome prediction (Spearman’s rank, ρ = 0.86) than the
ranking of features protein kinases to the whole genome prediction (Spearman’s rank, ρ = 0.51).
This finding suggests that the feature value distributions of TF WGD-duplicate and WGDsingletons are more similar to the genome at large. Therefore, the reason that TF duplicate
prediction model had lower performance was not simply because their feature values were
substantially different from other duplicate genes. Instead, the features examined simply have
lower importance in general for predicting TF retention (average MDA=11.3) than for other
genes (average MDA=47.9), suggesting there are additional features important for TF retention
that were not considered. For example, we might expect the number of DNA binding sites to be
predictive of duplication status as an indication of the breadth of function of the TF which is
related to the probability that a duplicate copy has been retained through subfunctionalization or
gene balance.

224

Furthermore, the most informative feature for classifying kinases and the whole genome,
the percent identity to the best matching paralog in A. thaliana, was less important when applied
to TFs (Table 3). Although the maximum percent identity of WGD-duplicates compared to
WGD-singletons is significantly higher in full genome (pv = 1e-320), protein kinases (pv = 1.1e36), and TFs (pv = 6.2e-12), the magnitude of the difference was greater for protein kinases
(11.2%) and the whole genome (11.3%) than TFs (4.4%). This is due to WGD duplicate TFs
having lower maximum percent identity (71.3%) than either kinases (75.2%, p=4.1e-24, t-test) or
all genes (72.5%, p=5.9e-83 , t-test), while WGD-singletons TF had higher identity (66.9%) than
kinases (64.0%, p=4.2E-35, t-test) and all genes (61.3%, p=1.9e-223, t-test). This observation
may related to non-duplicate TF genes having apparent paralogs more often than non-duplicate
genes do on average across the A. thaliana genome (Fig S2). The variance in the importance of
maximum percent identity accounts for most of the performance difference across the classifiers
as removing this feature yields similar results from all three (Fig S3). Similarly, inflating the
difference in the percent identity of TF WGD-duplicates and WGD-singletons from 4.4% to
11.2% (the difference for protein kinases) would raise the predicted retention of TF from the Îł
WGD from 2.50 to 2.94, making up for more than half of the original error.
We would expect that other features used in our linear models would also be useful for
classifying genes within function groups. However, the average importance rank of features
found in more than one linear model was low (13.9 of 20), with the maximum expression value
in RNA-seq being the worst feature in both the whole genome and TF classifiers. Of the four
linear model features, mean expression in AtGenExpress had the highest rank in the whole
genome (12th), TF (7th), and kinase classifiers (5th). However, the difference in mean
expression between WGD-duplicates and WGD-singletons was not consistent: WGD duplicates

225

genes were more highly expressed across the whole genome (+0.32, p=4.0e-23), and TFs (+0.37,
p=1.0e-4), but in protein kinases WGD-singletons were more highly expressed, though not at a
significant level (-1.1, p=0.77). Hence, not only does relationship between gene features and
retention depend on the gene function, but the relationship within individual function groups can
be the opposite direction of the relationship across function groups. For example, the high
retention of the TF function group is in part due to relatively low average expression in
AtGenExpress, but within TFs, genes with higher average expression are more often WGD
duplicates. This suggests that selection for duplicate retention is dependent not only on function
and features, but their interaction as well, though the exact nature of these interactions is beyond
the scope of this study.

226

Supplemental Figure 4.1 Frequency distribution of synonymous substitution rate (Ks)
between putative paralogs. Colors correspond to putative paralogs that are TFs (blue), kinases
(green), or any other genes (orange). Known WGD and tandem duplicates are excluded from
each group.

227

Supplemental Figure 4.2 Difference between the observed rates of duplicate retention and
rates predicted by the linear models of duplicate retention. Different events are indicated by
color (ι = orange, β = green, γ = blue). Positive values indicate the observed rate is larger than
the prediction while negative values indicate the observed rate is less than the prediction.

228

Supplemental Figure 4.3 Difference in expression quartile of individual TF duplicates
compared to their ancestral state. Expression subsets (Ctrl, LightDev, Diff, and Stress) are
indicated on the left and WGD event (ι = left, β = middle, γ = right) along the top. Heatmaps
show the z-scores of the observed frequency of each difference compared to the expected
frequency. Color correlates with the magnitude of the z-score, with darker red values indicating
counts further above random expectation and darker blue values indicating counts further below
random expectation.

229

Supplemental Figure 4.4 Deviation of pairs of TF WGD-duplicates from their ancestral
state. Deviation is defined as the difference value that each duplicate in a pair has from its
ancestral state. Expression subsets (Ctrl, LightDev, Diff, and Stress) are indicated on the left and
WGD event (ι = left, β = middle, γ = right) along the top. Heatmaps show the z-scores of the
observed frequency of the WGD-duplicate pair deviation compared to the expected frequency
across all three duplicate events (ι = top, β = middle, γ = bottom). Color correlates with the
magnitude of the z-score, with darker red values indicating counts further above random
expectation and darker blue values indicating counts further below random expectation.

230

Supplemental Figure 4.5 ODE models of the evolution of ancestral expression into either a
higher or lower expression quartile. In this model, we consider the transition of a single WGD
231

Supplemental Figure 4.5 (cont’d)
duplicate from an ancestral expression state (O) to either a higher (+) or lower (-) expression
state. Results for one (left column) and two (right column) parameter models show the change in
time (x-axis) of the frequency (y-axis) of each state (O = orange, + = blue, - = green). Curves
represent the continuous output of the model while symbols indicate the observed values on
which the models were built (O = circle, + = square, - = triangle).

232

Supplemental Figure 4.6 ODE models of TF WGD-duplicate expression evolution relative
to ancestral state for the Ctrl, Diff, and Stress expression subsets. In this model, we consider
233

Supplemental Figure 4.6 (cont’d)
the transition of the WGD-duplicate pair expression between three possible states relative to their
ancestral state (O = both retained, I = one retained, II = neither retained). Results for one (left
column) and two (right column) parameter models showing the change in time (x-axis) of the
frequency (y-axis) of each WGD-duplicate-pair state (O = orange, I = blue, II = green). Curves
represent the continuous output of the models while the symbols indicate the observed values on
which the models were built (O = circle, I = square, II = triangle).

234

Supplemental Table 4.1 Data sets used in linear model of duplicate retention
Name

Description

Source

Use1

Gene Count

Number of genes in each Group

Internal

Kept

Paralog Ks

Synonymous substitution rate relative to the highest

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Kept

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Dropped

Mean of expression in AtGenExpress data

Internal

Kept

Expression Mean

Mean of expression in StressTreatment subset of

Internal

Dropped

(Stress Data)

AtGenExpress Data

Expression Mean

Mean of expression in Development and Light subset of

Internal

Dropped

(DevLight Data)

AtGenExpress data

scoring BLAST hit Arabidopsis thaliana
Paralog Ka

Non-synonymous substitution rate relative to highest
scoring BLAST hit in Arabidopsis thaliana (derived from
paralog Ks and Ka/Ks)

Paralog Ka/Ks

Ka/Ks relative to the highest scoring BLAST hit
Arabidopsis thaliana

Ka/Ks (A. lyrata)

Median Ka/Ks relative to genes in Arabidopsis lyrata in
the same OrthoMCL Cluster

Ka/Ks (O. sativa)

Median Ka/Ks relative to genes in Oryza sativa in the
same OrthoMCL Cluster

Ka/Ks (P. patens)

Median Ka/Ks relative to genes in Physcomitrella
patensin the same OrthoMCL Cluster

Ka/Ks (P.

Median Ka/Ks relative to genes in Populus trichocarpa

trichocarpa)

in the same OrthoMCL Cluster

Ka/Ks (V. vinfera)

Median Ka/Ks relative to genes in A. lyrata in the same
OrthoMCL Cluster

Expression Mean
(AtGenExpress)

235

Supplemental Table 4.1 (cont’d)
Expression Mean

Mean of expression in Control subset of AtGenExpress

Internal

Dropped

(Control Data)

data

Expression Mean

Mean of expression in Stress Difference data set

Internal

Dropped

Median of gene expression in AtGenExpress

Lloyd et al. (2015)

Dropped

Expression

Median absolute deviation of expression over median

Lloyd et al. (2015)

Kept

MAD/Median

expression using AtGenExpress

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Kept

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Dropped

Internal

Dropped

Internal

Dropped

(Diff Data)
Expression Median
(AtGenExpress)

(AtGenExpress)
Expression Breadth

Number of AtGenExpress expression data sets with log2

(AtGenExpress)

intensity > 4

Expression

Max of expression correlation with genes in the same

Correlation

OrthoMCL cluster using data from AtGenExpress

(AtGenExpress)
Expression

Max of expression correlation with genes in the same

Correlation

OrthoMCL cluster that have Ks < 2 using data from

(AtGenExpress, Ks

AtGenExpress

< 2)
Expression Module

Size of co-expression module defined using K-means

Size

clustering of expression vectors from AtGeneExpress

(AtGenExpress)
Expression Breadth

Number of expression in RNA-Seq Data set where the

(RNASeq)

95% confidence interval of FPKM does not include 0

Expression Mean

Mean of expression in RNA-Seq Data Set

(RNASeq)

236

Supplemental Table 4.1 (cont’d)
Expression Median

Median of expression in RNA-Seq Data Set

Internal

Dropped

Maximum of expression in the RNA-Seq Data Set

Internal

Kept

Sequence

Percent identity of BLAST hits with 34 plant species

Lloyd et al. (2015)

Dropped

Conservation

(described in Lloyd et al., 2015)

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Kept

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Dropped

Lloyd et al. (2015)

Kept

(RNASeq)
Expression
Maximum
(RNASeq)

(Viridiplantae)
Sequence

Percent identity of BLAST hits with 8 fungal species

Conservation

(described in Lloyd et al., 2015)

(Fungi)
Sequence

Percent identity of BLAST hits with 8 metazoan species

Conservation

(described in Lloyd et al., 2015)

(Metazoa)
Function

Number of functional interactions annotated in AraNet

Interactions

(http://www.functionalnet.org/aranet/)

(AraNet)
Protein-Protein

Number of protein-protein interactions annotated in the

Interactions (AIMC)

Arabidopsis Interaction Network Map
(http://interactome.dfci.harvard.edu/A_thaliana/)

Nucleotide Diversity

Nucleotide Diversity (Pi) calculate between 80

(Pi)

Arabidopsis accessions

Number of Protein

Number of annotated protein domains

Lloyd et al. (2015)

Kept

Length of gene's longest protein product

Lloyd et al. (2015)

Kept

Domains
Protein Length (in
Amino Acids)

237

Supplemental Table 4.1 (cont’d)
Maximum Percent

Percent identity with the highest scoring BLAST hit in

Identity

Arabidopsis thaliana

Gene Family Size

Number of genes in the same OrthoMCL Family

Lloyd et al. (2015)

Kept

Lloyd et al. (2015)

Dropped

(OrthoMCL)

1. Indicates whether the feature was kept in the final linear model according to the filtering
procedures described in Methods

238

Supplemental Table 4.2 Subsets of AtGenExpress used for ancestral expression inference
Data Set

Experimental

Total

Number of Inferred States

Conditions

Samples

Control

4

24

13600

Light and

34

91

47792

Stress Treatment

32

70

37391

Stress Differential

48

175

66602

Development

239

Supplemental Table 4.3 Observed and expected frequency of duplicates TF pairs in a
conserved, partitioned, and diverged state
Data Set

Duplicate State

Îą WGD

Stress

Diff

LightDev

Ctrl

Observed

β WGD

Expected

Observed

Îł WGD

Expected

Observed

Expected

Conserved

0.54

0.58

0.46

0.50

0.49

0.49

Partitioned

0.45

0.37

0.49

0.42

0.42

0.42

Diverged

0.01

0.05

0.05

0.08

0.09

0.08

Conserved

0.53

0.58

0.47

0.51

0.49

0.50

Partitioned

0.45

0.37

0.48

0.41

0.42

0.42

Diverged

0.01

0.06

0.04

0.08

0.08

0.08

Conserved

0.57

0.62

0.53

0.57

0.49

0.51

Partitioned

0.43

0.34

0.44

0.39

0.43

0.43

Diverged

0.01

0.04

0.03

0.05

0.08

0.07

Conserved

0.54

0.59

0.48

0.52

0.46

0.47

Partitioned

0.44

0.36

0.47

0.41

0.44

0.44

Diverged

0.01

0.05

0.05

0.07

0.10

0.10

240

Supplemental Table 4.4 Experimental conditions used in each subset of AtGenExpress
Data Set
Controls

Conditions
Biotic control, Shoot control, Root control, Cell control

Light and

1st node, Carpels, Cauline leaves, Continuous blue light, Continuous darkness,

Development

Continuous far red light, Continuous red light, Continuous white light, Flower,
Flower, Hypocotyl, Leaf 7, Leaf 7, Leaves, Mature Pollen, Mature Rosette,
Mutant Rosette, Mutant Shoot, Pedicel, Petals, Red light pulse, Rif, Roots,
Rosette, Seedling, Seed siliques, Senescing leaves, Shoot apex, Sepals,
Stamens, Stem, UV-A-B light pulse, UV-A light pulse, Vegetative Rosette

Stress

avrRpm1, DC3000, Flg22, GST, GST-NPP1, H2O, HrcC, HrpZ, LPS, MgCl1,

Treatment

P, Psph, heat, MgCa, Root Cold, Root Drought, Root Genotoxic, Root Heat,
Root Osmotic, Root Oxidative, Root Salt, Root UV-B, Root Wounding, Shoot
Cold, Shoot Drought, Shoot Genotoxic, Shoot heat, Shoot Osmotic, Shoot
Oxidative, Shoot Salt, Shoot UV-B, Shoot Wounding

Stress

Root Cold 4C, Root Drought, Root Genotoxic, Root Heat, Root Osmotic, Root

Differential

Salt, Root UV-B, Root Wounding, Root avrRpm1, Root DC3000, Root Flg22,
Root GST-NPP1, Root HrcC, Root HrpZ, Root P. infestans, Root Psph, Root
Cold 4C, Root Columnar, Root Cortex, Root Drought, Root Endo, Root Epi,
Root Genotoxic, Root Heat, Root Osmotic, Root Oxidative, Root Protophl,
Root Salt, Root Stele, Root UV-B, Root Wounding, Shoot avrRpms1, Shoot
Cold4C, Shoot D3C300, Shoot Drought, Shoot Flg22, Shoot genotoxic, Shoot
GST-Npp1, Shoot Heat, Shoot HrcC, Shoot HrpZ, Shoot Osmotic, Shoot P.
infestans, Shoot Psph, Shoot Salt, Shoot UV-B, Shoot Wounding

241

Supplemental Table 4.5 RNA-seq data sets
Sequence Read Archive Identifiers
SRR1257404, SRR1257403, SRR1257402, SRR1257401, SRR1257392, SRR1257391,
SRR1257390, SRR1257389, SRR976397, SRR976398, SRR976391, SRR976394,
SRR929001, SRR929000, SRR921316, SRR921315, SRR921314, SRR921313,
SRR921312, SRR921311, SRR671949, SRR671948, SRR671947, SRR671946,
SRR653578, SRR653577, SRR653576, SRR653575, SRR653574, SRR653573,
SRR653572, SRR653571, SRR653570, SRR653569, SRR653568, SRR653567,
SRR653566, SRR653565, SRR653564, SRR653563, SRR653562, SRR653561,
SRR653557, SRR653556, SRR653555, SRR649539, SRR649538, SRR649537,
SRR634971, SRR634970, SRR634969, SRR584126, SRR584125, SRR584120,
SRR584119, SRR520239, SRR520238, SRR520237, SRR515492, SRR515491,
SRR515490, SRR515489, SRR479032, SRR477076, SRR477075, SRR452279,
SRR452278, SRR452277, SRR452275, SRR452276, SRR452274, SRR445738,
SRR445737, SRR441559, SRR441558, SRR402997, SRR402996, SRR402995,
SRR402994, SRR391052, SRR391051, SRR070570, SRR070571, SRR069568,
SRR069569, SRR069565, SRR069566, SRR069567, SRR069558, SRR069559,
SRR069556, SRR069557, SRR974753, SRR974752, SRR974751, SRR974750,
SRR652153, SRR652152, SRR652151, SRR652150, ERR274309, ERR274308,
ERR274311, ERR274310, SRR1146545, SRR1055106, SRR1023821, SRR1020622,
SRR1020621, SRR1005386, SRR1005385, SRR1005239, SRR1005238, SRR1001910,
SRR1001909, SRR902025, SRR835483, SRR800754, SRR800753, SRR609268,
SRR609267, SRR578948, SRR578947, SRR522012, SRR520363, SRR520364,
SRR505746, SRR505745, SRR505744, SRR505743, SRR505137, SRR505135,
SRR493043, SRR493039, SRR493036, SRR493033, SRR445736, SRR445735,
SRR445214, SRR445215, SRR445216, SRR445217, SRR445218, SRR445209,
SRR445210, SRR445211, SRR445212, SRR445213, SRR445204, SRR445205,
SRR445206, SRR445207, SRR445208, SRR443169, SRR443165, SRR443164,
SRR443163, SRR419186, SRR419182, SRR404277, SRR403907, SRR403903,
SRR402370, SRR402371, SRR390314, SRR390312, SRR390313, SRR390311,
SRR390310, SRR390308, SRR390309, SRR390306, SRR390307, SRR390305,
SRR390303, SRR390304, SRR390302, SRR388670, SRR388669, SRR388668,
SRR970149, SRR953401, SRR953400, SRR953399, SRR952321, SRR847505,
SRR847506, SRR847503, SRR847504, SRR847501, SRR847502, SRR218098,
SRR522916, SRR360147, SRR360152, SRR360153, SRR360154, SRR360205,
SRR218099, SRR218100, SRR218092, SRR339951, SRR218101, SRR218102,
SRR218096, SRR218097, SRR218089, SRR218090, SRR218085, SRR218086,
SRR218087, SRR218088, SRR218094, SRR218095

242

Supplemental Table 4.6 Genes belonging to each GO-defined function group
Function
Group
ATP Binding
(GO:0005524)

Genes
AT1G14390, AT1G58050, AT1G01910, AT3G24240, AT2G24130, AT3G52570,
AT5G26860, AT4G36180, AT4G17380, AT1G25320, AT5G49030, AT3G59760,
AT5G44800, AT3G45300, AT2G23300, AT3G10350, AT3G57300, AT4G22730,
AT4G01800, AT5G57450, AT4G36270, AT2G07040, AT5G63410, AT1G05120,
AT4G02930, AT4G37870, AT4G36290, AT2G19860, AT3G10690, AT1G11100,
AT2G26730, AT5G56030, AT2G27490, AT2G02090, AT5G58150, AT1G10850,
AT2G45500, AT1G62750, AT5G17760, AT3G05780, AT1G32060, AT3G24340,
AT1G74310, AT1G17750, AT3G28450, AT4G01020, AT5G51560, AT1G17290,
AT2G42290, AT3G24660, AT4G23740, AT1G24290, AT2G28970, AT1G01220,
AT5G38830, AT2G25790, AT3G46370, AT4G21800, AT1G70460, AT1G28440,
AT2G21450, AT3G52200, AT5G21326, AT2G27600, AT1G62950, AT2G27170,
AT3G28520, AT1G29900, AT5G65710, AT5G06820, AT3G53230, AT1G26190,
AT1G49250, AT3G14350, AT3G20190, AT5G07660, AT3G49670, AT5G66760,
AT2G13370, AT5G61030, AT1G74260, AT5G49770, AT3G56100, AT5G46330,
AT3G19210, AT3G57760, AT1G75640, AT1G18130, AT2G15300, AT1G69990,
AT4G00570, AT2G13800, AT4G24190, AT4G13850, AT3G27440, AT4G35740,
AT5G03290, AT5G56040, AT4G34220, AT1G48650, AT3G54660, AT5G22370,
AT5G40000, AT5G17730, AT3G62120, AT3G09660, AT5G63950, AT2G24230,
AT4G25120, AT2G16250, AT4G28650, AT3G27190, AT2G26080, AT3G24495,
AT4G10320, AT1G48480, AT3G13065, AT2G20420, AT1G74230, AT5G45840,
AT1G74330, AT3G50940, AT1G58060, AT5G67520, AT5G14610, AT5G07810,
AT5G61480, AT1G51980, AT4G26300, AT4G31180, AT2G32800, AT5G14470,
AT2G02220, AT3G06480, AT3G06483, AT3G57640, AT2G30800, AT1G29750,
AT5G04110, AT4G23240, AT4G18640, AT3G02130, AT5G10020, AT5G46280,
AT3G18810, AT3G28610, AT5G16715, AT2G35120, AT3G51740, AT4G36280,
AT5G65720, AT4G12790, AT5G56000, AT4G01900, AT3G20475, AT1G27880,
AT1G63940, AT1G72180, AT4G23900, AT3G56370, AT2G26280, AT1G27190,
AT1G73080, AT3G14840, AT5G61460, AT5G35390, AT1G78900, AT5G07440,
AT4G37840, AT2G45280, AT2G36570, AT5G19310, AT5G52520, AT1G66830,
AT1G53730, AT4G03390, AT2G25840, AT1G29870, AT2G27060, AT5G60730,
AT1G72300, AT2G18470, AT5G48600, AT2G44350, AT3G16600, AT3G02880,
AT1G09970, AT3G47110, AT5G44635, AT3G50930, AT2G01130, AT1G06840,
AT5G20690, AT1G53420, AT5G24100, AT5G45800, AT1G14610, AT2G44980,
AT3G50230, AT3G10270, AT3G13170, AT3G53590, AT1G14000, AT4G11010,
AT3G22880, AT4G00960, AT5G44700, AT5G26742, AT1G64210, AT5G16050,
AT2G41820, AT3G02660, AT4G23270, AT4G20270, AT1G07200, AT3G02065,
AT1G51830, AT2G01950, AT1G79930, AT1G50410, AT4G04350, AT5G50920,
AT1G72040, KATE, AT1G75820, AT2G01460, AT1G50480, AT5G22010, AT3G23890,
AT5G01890, AT4G37250, AT5G63310, AT5G59660, AT1G55810, AT3G20010,
AT3G54670, AT1G63680, AT1G51390, AT3G19700, AT5G58300, AT2G33840,
AT3G56300, AT1G79620, AT3G06010, AT3G08680, AT3G28570, AT3G28580,
AT2G01210, AT4G33760, AT3G42850, AT5G01950, AT4G35030, AT3G49240,
AT1G73980, AT3G04600, AT3G47090, AT5G37450, AT1G72460, AT2G02780,
AT4G12060, AT2G14050, AT2G14750, AT3G25840, AT5G14210, AT1G09620,
AT3G54280, AT4G35520, AT3G42670, AT3G45450, AT5G49780, AT5G41180,
AT3G47570, AT2G13560, AT4G31250, AT5G20480, AT3G28600, AT5G16590,
AT4G37910, AT4G24830, AT1G65190, AT3G45770, AT5G65700, AT1G21650,
AT3G06580, AT1G31420, AT1G43910, AT5G56010, AT1G78980, AT5G05130,
AT3G23990, AT1G66530, AT2G18190, AT1G05910, AT2G18193, AT3G07770,
AT1G48310, AT1G03030, AT2G31170, AT5G55200, AT5G04895, AT3G24320,
AT2G18760, AT2G16440, AT3G12580, AT5G54590, AT5G51350, AT4G30250,

243

Supplemental Table 4.6 (cont’d)

catalytic activity
(GO:0003824)

AT3G03770, AT1G74360, AT3G20040, AT2G16390, AT4G25370, AT5G47040,
AT2G45590, AT4G20140, AT2G35920, AT3G28540, AT5G10880, AT2G04030,
AT2G45340, AT5G17740, AT3G48870, AT1G72330, AT1G49270, AT1G63990,
AT5G51070, AT3G24550, AT1G17410, AT1G68400, AT3G57830, AT2G25140,
AT1G08600, AT4G29130, AT5G09590, AT3G47580, AT3G59410, AT1G08130,
AT4G28490, AT3G13490, AT5G63710, AT3G48000, AT1G02670, AT1G67840,
AT5G45780, AT2G33170, AT3G17840, AT1G07190, AT5G15920, AT1G33390,
AT3G10700, AT3G03900, AT5G53320, AT2G46020, AT1G44900, AT1G50460,
AT4G34200, AT3G12810, AT2G05710, AT5G05160, AT4G24280, AT5G54090,
AT1G67510, AT3G05790, AT4G39940, AT5G48940, AT2G45490, AT1G22300,
AT5G64580, AT4G16130, AT1G35710, AT2G39730, AT5G15450, AT4G14350,
AT1G52290, AT2G07690, AT2G32850, AT1G35720, AT3G44740, AT5G06580,
AT5G22750, AT1G56130, AT3G27730, AT2G33210, AT3G28040, AT2G46370,
AT5G20420, AT4G02460, AT2G46620, AT5G19720, AT4G26510, AT5G04130,
AT3G18524, AT1G66730, AT5G58720, AT1G28350, AT4G09320, AT1G61140,
AT5G65690, AT5G40870, AT1G63430, AT5G43020, AT3G11710, AT3G29800,
AT5G10370, AT4G29990, AT1G34110, AT2G26700, AT4G02060, AT4G39280,
AT5G43530, AT3G01640, AT1G12460, AT3G55010, AT1G07650, AT1G50610,
AT1G48030, AT5G08670, AT3G55400, AT5G63930, AT1G65070, AT5G50780,
AT5G02820, AT1G47840, AT5G18170, AT3G42880, AT2G34560, AT4G39270,
AT5G67200, AT1G45332, AT3G17240, AT5G38480, AT4G29380, AT2G31880,
AT5G40010, AT3G58140, AT3G26560, AT5G53890, AT4G08920, AT1G60630,
AT5G20040, AT5G17750, AT5G67280
AT1G14290, AT1G64660, AT1G78050, AT1G50090, AT4G32790, AT5G20260,
AT5G40270, AT3G55180, AT3G14790, AT4G12960, AT2G26000, AT2G31955,
AT3G03990, AT2G04440, AT3G46440, AT2G38660, AT1G02270, AT4G02850,
AT1G03210, AT5G37000, AT1G74290, AT5G19290, AT5G65280, AT5G62220,
AT3G10572, AT4G10100, AT2G47630, AT4G14440, AT3G17365, AT5G44480,
AT3G20650, AT2G35100, AT3G49680, AT3G55190, AT3G23820, AT3G62860,
AT3G23940, AT1G08940, AT5G02970, AT4G12870, ARA2, AT3G26820, AT5G41250,
AT4G34360, AT4G13360, AT4G29530, AT1G78500, AT1G29840, AT3G24030,
AT2G29630, ARA1, AT5G03800, AT5G41650, AT2G43400, AT4G00620, AT4G22330,
AT5G40290, AT4G14430, AT2G47760, AT5G57040, AT5G11130, AT1G12350,
AT3G19710, AT4G12250, AT5G14180, AT2G25100, AT4G38800, AT3G10690,
AT5G48010, AT1G53500, AT5G01260, AT1G13635, AT4G08170, AT3G03230,
AT5G27410, AT1G63450, AT4G30440, AT2G20370, AT1G30620, AT3G14890,
AT4G34700, AT3G05170, AT5G37530, AT1G76730, AT1G07645, AT1G27440,
AT1G64185, AT2G22570, AT5G14980, AT4G33540, AT4G30540, AT2G01730,
AT4G24340, AT3G57630, AT4G18270, AT1G10060, AT1G34380, AT5G38360,
AT5G23220, AT2G32410, AT2G29040, AT2G45310, AT2G34770, AT1G01290,
AT4G22756, AT5G57800, AT2G39725, AT4G33180, AT2G46370, AT1G13700,
AT5G52810, AT1G25375, AT4G00110, AT1G22170, AT1G09935, AT1G74300,
AT3G16700, AT4G12230, AT1G34270, AT5G64150, AT3G12290, AT5G19670,
AT3G47560, AT1G68470, AT4G22580, AT5G11910, AT3G04390, AT4G16690,
AT1G10070, AT4G25720, AT1G74680, AT4G13990, AT3G53520, AT2G25710,
AT3G62830, AT3G13800, AT5G25820, AT2G28760, AT4G16210, AT3G26780,
AT2G43980, AT5G16890, AT2G31990, AT5G04120, AT4G12890, AT1G65520,
AT1G69640, AT5G26570, AT3G12260, AT3G24730, AT2G17280, AT2G47650,
AT1G05350, AT4G22753, AT5G36150, AT2G20770, AT3G60510, AT2G39420,
AT3G14990, AT1G48420, AT5G41120, AT2G23820, AT5G11610, AT1G37150,
AT5G49570, AT5G62840, AT4G25434, AT3G45400, AT1G26160, AT3G42180,
AT5G02080, AT5G08290, AT4G02860, AT4G15940, AT5G01580, AT5G22940,
AT3G19820, AT1G74260, AT5G44930, AT2G32960, AT1G54570, AT4G38370,

244

Supplemental Table 4.6 (cont’d)

defense response
(GO:0006952)

AT2G34850, AT1G02190, AT5G22460, AT4G30530, AT3G07620, AT3G47590,
AT1G67410, AT3G60910, AT4G30550, AT1G07080, AT5G25310, AT3G52050,
AT1G17890, AT3G26840, AT5G11560, AT5G65780, AT1G78570, AT1G52160,
AT1G76060, AT4G12900, AT3G03240, AT3G50520, AT5G61840, AT5G33290,
AT1G52920, AT4G29120, AT4G38040, AT4G00600, AT2G32740, AT2G42160,
AT5G59290, AT3G49360, AT5G63420, AT3G16260, AT4G20870, AT3G62810,
AT1G34340, AT3G16190, AT2G37700, AT5G57850, AT4G15370, AT5G28840,
AT3G58830, AT5G24400, AT5G41130, AT2G19550, AT5G11350, AT3G03650,
AT5G23230, AT3G05190, AT1G12850, AT4G12110, AT1G08310, AT1G02000,
AT4G20460, AT1G21480, AT5G61220, AT2G32750, AT5G22620, AT5G42600,
AT1G50110, AT4G37500, AT5G16760
AT3G44480, AT5G06870, AT1G53350, AT5G44870, AT3G50950, AT5G51060,
AT5G51700, AT1G12290, AT1G19610, AT1G61100, AT1G72840, AT4G04110,
AT1G72920, AT1G64070, AT1G55010, AT4G02600, AT1G14410, AT5G25910,
AT2G03300, AT3G46530, AT1G57830, AT2G39940, AT1G52040, AT1G61070,
AT4G09360, AT3G05370, AT5G48620, AT4G16920, AT5G66910, AT1G69550,
AT2G26380, AT2G15170, AT5G46470, AT5G17890, AT5G46270, AT5G56030,
AT1G66100, AT5G63660, AT3G23180, AT2G32660, AT1G17615, AT1G63870,
AT1G26700, AT4G11210, AT1G61310, AT3G26830, AT3G05650, AT5G23820,
AT5G38340, AT3G51570, AT1G63750, AT1G57650, AT1G58400, AT4G36140,
AT4G19510, AT2G38900, AT1G52660, AT1G09090, AT4G10780, AT5G18360,
AT1G33560, AT4G08450, AT4G03550, AT1G72870, AT2G43510, AT5G43740,
AT3G11820, AT4G02150, AT2G02100, AT5G45080, AT2G39200, AT4G13920,
AT1G66090, AT1G12220, AT1G64160, AT2G33050, AT5G24780, AT5G41740,
AT1G42560, AT3G24900, AT2G02740, AT2G23960, AT3G04220, AT3G13662,
AT3G13660, AT1G79680, AT3G23120, AT2G32140, AT3G25010, AT5G64930,
AT1G75830, AT5G44420, AT4G16890, AT1G59124, AT5G13160, AT5G18350,
AT2G17430, AT4G16960, AT2G34930, AT1G71260, AT5G35450, AT1G12210,
AT2G22330, AT1G61180, AT5G45260, AT5G45060, AT3G05660, AT2G03760,
AT2G33670, AT3G46860, AT5G06860, AT5G44900, AT1G63360, AT4G09430,
AT5G36930, AT1G58848, AT3G14460, AT3G46730, AT3G15010, AT1G72850,
AT1G64060, AT1G72910, AT1G72520, AT3G49120, AT5G46760, AT3G51560,
AT4G14370, AT3G45290, AT4G24230, AT3G53260, AT4G33300, AT3G09710,
AT2G43520, AT4G23515, AT4G11170, AT1G66340, AT5G65970, AT3G05360,
AT4G16930, AT5G17880, AT1G47890, AT2G02130, RPP22, AT4G13810, AT4G11190,
AT5G46260, AT5G48780, AT2G02140, AT2G43910, AT1G17600, AT1G58390,
AT5G04230, AT2G15010, AT3G23010, AT1G61300, AT1G63740, AT3G25020,
AT5G44510, AT1G56520, AT2G15080, AT1G52900, AT5G38330, AT5G04720,
AT1G58410, AT1G58807, AT4G19500, AT4G17880, AT5G42510, AT5G49040,
AT4G04220, AT3G44630, AT1G60320, AT2G03030, AT3G50020, AT4G16940,
AT2G14080, AT5G46520, AT3G20600, AT1G72260, AT1G17420, AT4G24250,
AT2G15220, AT3G25510, AT1G65870, AT3G52450, AT1G19230, AT3G14470,
AT1G50180, AT5G45090, AT2G33060, AT4G09420, AT4G38700, AT3G04210,
AT3G23110, AT3G13650, AT5G46450, AT1G59780, AT1G65390, AT5G64905,
AT4G26090, AT5G43580, AT5G17680, AT2G21100, AT5G07390, AT2G47730,
AT5G05170, AT3G11080, AT5G45230, AT3G61220, AT5G40170, AT5G11250,
AT1G72950, AT4G39950, AT3G15700, AT5G45070, AT3G56860, AT1G58602,
AT1G56510, AT5G15130, AT2G35930, AT5G44910, AT1G58170, AT1G51480,
AT5G40100, AT5G43730, AT5G41540, AT1G72860, AT1G72900, AT5G45200,
AT4G39030, AT4G30070, AT5G48770, AT3G49110, AT4G16990, AT5G47910,
AT5G40910, AT1G22900, AT4G19920, EDS9, AT4G19925, AT4G16900, AT1G63350,

245

Supplemental Table 4.6 (cont’d)

DNA
endoreduplication
(GO:0042023)

hydrolase activity
hydrolyzing Oglycosyl
compounds
(GO:0004553)

AT1G15890, AT5G46490, AT5G58120, AT2G02120, AT4G11340, AT1G31540,
AT4G11180, AT2G30860, AT3G48090, AT3G23240, AT5G47250, AT2G16870,
AT1G27180, AT4G27190, AT1G63880, AT1G63730, AT1G45616, AT1G55210,
AT5G53760, AT1G57670, AT4G19530, AT3G20820, AT3G44670, AT5G46510,
AT1G72940, AT5G45240, AT1G09665, AT2G17050, AT4G26740, AT4G23570,
AT3G11010, AT3G11480, AT1G57850, AT1G12280, AT3G45860, AT1G72930,
AT2G41060, AT5G45220, AT3G16720, AT1G31580, AT1G10920, AT1G11000,
AT2G32680, AT3G44400, AT1G47370, AT5G41550, AT4G19910, VET1, AT4G16950,
AT5G66900, AT1G59620, AT2G23970, AT2G26010, AT5G43470, AT3G11340,
AT2G21110, AT5G43570, AT1G57630, AT4G23310, AT5G47260, AT1G66980,
AT5G18370, AT1G27170, AT2G14610, AT2G43710, AT1G59218, AT5G38350,
AT5G36910, AT3G55230, AT5G22690, AT3G46710, AT5G51630, AT3G11840,
AT5G45210, AT4G12010, AT1G61560, AT1G55020, AT5G49140, AT1G61190,
AT2G17060, AT5G38850, AT1G73050, CIR3, AT1G72890, AT5G66890, AT1G33590,
AT5G23400, AT4G23690, AT5G27060, AT4G23510, AT5G42500, AT2G37040,
AT1G52030, AT3G07040, AT5G42650, AT1G71400, AT1G65850, AT5G44920,
AT1G71390, AT4G23280, AT2G44110, AT1G12663, AT1G12660, AT5G55240,
AT5G17970, AT5G13530, AT5G41750, AT4G27220, AT5G63020, AT4G36150,
AT2G15130, AT2G26020, AT5G45000, AT5G47280, AT5G44430, AT5G15410,
AT5G05400, AT1G56540, AT1G62630, AT4G16860, AT5G11270, AT4G19520,
AT2G17480, AT5G45250, AT1G11310
AT3G21860, AT1G15570, AT2G27960, AT5G04470, AT2G19330, AT5G20570,
AT3G20780, AT5G42190, AT1G20930, AT2G20140, AT4G26760, AT3G53970,
AT3G08690, AT2G21550, AT1G49910, AT1G80370, AT1G70210, AT3G50070,
AT4G05190, AT1G64520, AT1G47870, AT5G24630, AT2G22490, AT2G23430,
AT1G69690, AT1G78770, AT1G77390, AT5G24330, AT3G13550, AT5G22220,
AT1G75950, AT5G27620, AT4G34160, AT1G50490, AT3G11270, AT1G03780,
AT3G48150, AT5G05560, AT5G11300, AT1G73690, AT1G66750, AT1G76540,
AT5G48820, AT3G60010, AT5G41700, AT2G40550, AT5G57950, AT1G15660,
AT2G16740, AT4G24820, AT4G28980, AT3G48750, AT3G12280, AT3G19150,
AT5G10440, AT3G54180, AT1G64230, RFI, AT5G03415, AT1G20200, AT5G02470,
AT3G42830, AT2G32710, AT3G24810, AT5G63610, AT3G25980, AT5G56150,
AT3G20060, AT2G18290, AT2G42260, AT4G22910, AT2G39090, AT5G08550,
AT2G03430, AT3G11520, AT3G15180, AT3G21850, AT3G48160, AT1G59540,
AT5G11510, AT1G75990, AT1G47230, AT4G11920, AT5G64760, AT1G48380,
AT5G13840, AT3G50630, AT5G09900, AT4G03270, AT5G51600, AT4G22970,
AT2G27970, AT2G20000, AT5G65420, AT1G49620, AT4G29040, AT3G16320,
AT4G14150, AT1G02970, AT1G06590, AT5G05780, AT1G29150, AT4G37630,
AT2G36010, AT5G25380, AT1G50240, AT1G18040, AT3G60840, AT3G59550,
AT4G38600, AT5G67100, AT3G19590
AT1G11820, AT4G02290, AT4G27830, AT3G55260, AT4G19820, AT1G77780,
AT3G57270, AT5G20870, AT3G60130, AT1G51470, AT1G75940, AT1G66270,
AT5G58480, AT4G23560, AT5G16580, AT2G44480, AT3G09260, AT4G29360,
AT1G64390, AT4G27820, AT4G19730, AT1G51490, AT1G75680, AT5G20940,
AT5G09730, AT3G57260, AT5G20250, AT5G48375, AT2G05790, AT3G18080,
AT3G62740, AT5G58090, AT3G47000, AT5G11920, AT3G60140, AT4G33810,
AT3G13790, AT1G22880, AT4G22100, AT3G57520, AT1G33220, AT4G33830,
AT4G33860, AT3G23770, AT5G20950, AT5G25980, AT2G20680, AT1G52400,
AT3G30540, AT3G62750, AT3G47040, AT3G47010, AT2G44470, AT5G11720,
AT2G19440, AT1G10050, AT4G19770, AT1G02310, AT1G55120, AT5G20390,
AT4G08160, AT4G38650, AT3G19620, AT3G03640, AT2G14690, AT5G28510,
AT3G43860, AT1G65610, AT2G01630, AT3G57240, AT3G55780, AT1G02640,
AT3G04010, AT5G63840, AT3G55430, AT4G11050, AT1G12240, AT3G10900,

246

Supplemental Table 4.6 (cont’d)

kinase activity
(GO:0016301)

AT3G45940, AT2G44460, AT5G67460, AT1G70710, AT2G32990, AT1G66280,
AT1G62660, AT4G19760, AT1G48930, AT2G39640, AT2G44490, AT5G42260,
AT3G13784, AT1G13130, AT1G55740, AT1G71380, AT2G44550, AT4G18340,
AT1G02850, AT4G17180, AT5G20340, AT5G49360, AT4G33840, AT3G23640,
AT1G09010, AT5G64570, AT5G55180, AT3G61810, AT3G18070, AT4G26830,
AT4G19750, AT1G30080, AT3G15800, AT4G21760, AT4G34480, AT4G19810,
AT2G44540, AT3G26140, AT5G18220, AT4G39000, AT5G24540, AT1G78060,
AT1G26560, AT1G60090, AT1G32860, AT3G07320, AT3G52600, AT3G46570,
AT2G27500, AT4G19740, AT2G44450, AT5G36890, AT1G47600, AT4G31140,
AT5G20330, AT1G05590, AT5G56590, AT4G01040, AT2G44570, AT5G16700,
AT4G16260, AT1G61810, AT5G64790, AT4G33820, AT1G64760, AT3G54440,
AT5G24550, AT5G49720, AT3G24330, AT4G14080, AT3G62710, AT4G09740,
AT3G47050, AT2G25630, AT4G01970, AT2G36190, AT3G13560, AT2G16230,
AT1G61820, AT2G44560, AT1G66250, AT1G23210, AT5G26000, AT1G02800,
AT3G06510, AT5G20560, AT1G77790, AT4G38300, AT4G39010, AT5G01930,
AT1G58370, AT5G54570, AT4G28320, AT3G10890, AT4G24260, AT5G44640,
AT5G10560, AT1G68560, AT5G04885, AT5G40390, AT3G21370, AT1G19940,
AT2G26600, AT4G19800, AT4G19720, AT5G24090, AT5G42100, AT5G66460,
AT3G60120, AT5G42720, AT3G26130, AT2G32860, AT5G17500
AT4G40010, AT5G46080, AT1G51660, AT5G66850, AT2G25880, AT5G02290,
AT1G01540, AT1G73500, AT5G27510, AT1G28390, AT3G17750, AT1G68400,
AT1G67470, AT3G59350, AT2G02800, AT4G36180, AT4G10390, AT1G17160,
AT1G25320, AT4G28860, AT1G09600, AT2G28940, AT5G15730, AT2G39180,
AT3G48260, AT3G23340, AT1G29730, AT4G04500, AT2G23300, AT3G45430,
AT4G28540, AT5G02070, AT3G57710, AT3G26940, AT3G63280, AT1G65250,
AT4G23250, AT1G29230, AT4G23150, AT4G01330, AT1G12580, AT1G71530,
AT5G08590, AT2G25760, AT2G14510, AT5G07620, AT5G10930, AT3G15220,
AT2G21480, AT1G16440, AT5G59270, AT5G56580, AT1G03930, AT4G38230,
AT1G51890, AT1G47890, AT5G65500, AT4G04570, AT2G24360, AT3G46330,
AT2G41970, AT2G40560, AT3G53380, AT2G18530, AT2G40120, AT5G22050,
AT4G33950, AT4G24480, AT2G43230, AT2G32660, AT4G23320, AT5G67080,
AT3G12690, AT5G60310, AT2G34180, AT5G57035, AT4G38830, AT4G14780,
AT3G13670, AT2G17220, AT3G23000, AT3G54180, AT1G22720, AT1G54610,
AT5G10520, AT3G28450, AT3G05650, AT1G11350, AT1G09000, AT5G38990,
AT2G42290, AT1G77720, AT1G70520, AT4G23740, AT4G13020, AT2G01460,
AT5G04510, AT1G73460, AT2G28970, AT5G49470, AT3G50730, AT2G46700,
AT1G16130, AT3G44200, AT2G25790, AT1G67000, AT1G56120, AT3G46370,
AT3G44610, AT1G28440, AT1G67580, AT1G66920, AT3G19100, AT1G61550,
AT5G66710, AT2G39110, AT1G62950, AT1G79670, AT5G60080, AT1G61380,
AT4G39110, AT1G64630, AT2G17170, AT4G01370, AT1G33560, AT4G16970,
AT5G61350, AT5G06820, AT5G66880, AT5G47850, AT1G26190, AT4G28880,
AT3G01085, AT3G45670, AT3G20200, AT5G40540, AT1G67720, AT2G37050,
AT1G53430, AT3G20190, AT1G06390, AT2G31390, AT4G28350, AT3G05050,
AT1G61420, AT4G26610, AT4G08850, AT3G45790, AT3G01490, AT2G07020,
AT3G10540, AT5G62230, AT1G33260, AT5G46330, AT4G13920, AT1G77280,
AT1G11410, AT3G48750, AT5G18610, AT2G41930, AT5G35980, AT3G28040,
AT2G33050, AT4G13190, AT1G52310, AT5G25910, AT1G75640, AT4G23160,
AT2G15300, AT1G69990, AT1G14370, AT4G11480, AT4G02630, AT2G42960,
AT1G51805, AT2G23950, AT1G79680, AT3G59790, AT4G33080, AT3G23120,
AT2G33580, AT5G65530, AT2G45910, AT5G59650, AT3G24660, AT5G51830,
AT3G20530, AT3G27440, AT4G34500, AT1G21230, AT5G56040, AT1G16120,
AT1G18350, AT2G40500, AT5G41730, AT5G39440, AT2G30360, AT3G07980,
AT5G13160, AT1G62400, AT5G54380, AT3G17410, AT1G70740, AT3G54030,

247

Supplemental Table 4.6 (cont’d)
AT5G60280, AT3G17840, AT1G20930, AT3G59480, AT2G16750, AT2G19230,
AT2G37840, AT4G31110, AT1G61610, AT1G69270, AT3G27190, AT4G24100,
AT1G24030, AT4G14580, AT4G23290, AT1G09440, AT3G17420, AT3G13065,
AT1G78940, AT3G46930, AT4G04695, AT1G11330, AT1G61860, AT1G51810,
AT1G73450, AT1G49350, AT5G01810, AT5G61550, AT1G53570, AT5G19450,
AT3G04910, AT1G18670, AT1G61460, AT1G61500, AT1G19600, AT4G32830,
AT2G04300, AT5G18190, AT1G16670, AT2G34290, AT4G08470, AT2G28930,
AT1G61480, AT3G46160, AT2G46070, AT3G45440, AT3G22750, AT4G11530,
AT2G05940, AT1G29750, AT1G32320, AT2G34650, AT3G46420, AT3G02130,
AT1G51790, AT5G48380, AT5G07140, AT5G13290, AT4G21230, AT4G35310,
AT2G14440, AT5G28290, AT3G05360, AT5G59260, AT5G67380, AT3G46400,
AT1G03920, AT5G58520, AT3G51740, AT1G07870, AT3G21450, AT1G07150,
AT4G18700, AT1G51800, AT5G01560, AT2G19190, AT3G59110, AT1G01450,
AT1G18390, AT3G27560, AT3G05370, AT5G60550, AT5G55560, AT5G28680,
AT2G41860, AT4G23230, AT3G13530, AT5G58140, AT1G23540, AT5G60890,
AT1G72180, AT5G01060, AT4G11900, AT1G07570, AT4G04960, AT1G73670,
AT1G18890, AT5G25110, AT5G58350, AT5G60320, AT2G43850, AT3G56370,
AT1G27190, AT5G18910, AT4G04700, AT1G51870, AT2G23070, AT5G45430,
AT5G24360, AT3G53640, AT2G20850, AT5G02800, AT5G11850, AT1G07880,
AT4G04490, AT2G35620, AT2G38620, AT4G19110, AT1G55610, AT4G02410,
AT3G50500, AT2G15080, AT3G12200, AT2G30980, AT1G21240, AT1G49730,
AT5G12480, AT5G39000, AT3G25010, AT2G28960, AT2G28990, AT3G20830,
AT3G50720, AT1G35670, AT1G80870, AT4G26890, AT1G66930, AT3G04810,
AT3G55550, AT1G66830, AT1G03740, AT2G19470, AT1G25390, AT3G58640,
AT4G21400, AT4G04220, AT2G38910, AT4G03390, AT5G11020, AT3G45860,
AT1G69200, AT5G16900, AT4G18950, AT4G29050, AT3G05140, AT3G59730,
AT2G38490, AT1G66750, AT5G35960, AT2G40270, AT3G55950, AT5G08160,
AT1G05100, AT3G21630, AT2G39360, AT3G02880, AT3G11870, AT1G34210,
AT1G09970, AT3G47110, AT1G17540, AT4G10730, AT3G45640, AT3G24540,
AT5G28080, AT5G20690, AT1G53420, AT1G60800, AT1G61430, AT4G32660,
AT3G53840, AT4G21940, AT3G02810, AT3G50230, AT3G59420, AT5G49760,
AT5G51270, AT5G25930, AT3G45780, AT2G23450, AT4G11460, AT3G06230,
AT5G59670, AT1G16260, AT1G06020, AT5G43910, AT1G12680, AT3G45410,
AT2G36570, AT1G54960, AT5G60300, AT1G01140, AT2G40580, AT3G17850,
AT2G41820, AT2G33060, AT4G11490, AT2G32510, AT1G18040, AT4G26540,
AT1G65800, AT5G14720, AT1G10940, AT1G08720, AT1G48260, AT1G51830,
AT4G13000, AT4G14340, AT2G43790, AT3G23110, AT5G62310, AT1G78290,
AT3G22420, AT1G21590, AT3G09830, AT2G23030, AT1G64300, AT3G08870,
AT3G57750, AT4G24740, AT2G30940, AT2G42550, AT1G63700, AT1G75820,
AT4G31170, AT5G24080, AT5G04870, AT1G53050, AT5G67520, AT2G43700,
AT3G11080, AT5G01890, AT4G37250, AT3G06620, AT4G23650, AT1G55810,
AT1G16760, AT5G24430, AT5G40170, AT3G17510, AT5G58300, AT5G16000,
AT1G66970, AT5G47070, AT4G18250, AT2G25440, AT3G13690, AT3G59830,
AT1G70430, AT3G55450, AT4G26690, AT5G63940, AT1G79620, AT3G08680,
AT3G58760, AT4G31100, AT3G51990, AT4G28706, AT1G61370, AT2G01210,
AT3G53570, AT5G38560, AT4G35030, AT4G10010, AT1G49160, AT2G29220,
AT2G17530, AT1G73980, AT3G51550, AT4G10260, AT1G16110, AT5G11400,
AT1G60630, AT1G01560, AT5G37450, AT1G72460, AT5G18500, AT5G39030,
AT3G20860, AT3G15890, AT3G62220, AT3G25490, AT3G09010, AT1G49180,
AT5G58950, AT1G22870, AT5G19360, AT3G28690, AT3G25250, AT5G01820,
AT5G41260, AT4G20450, AT1G29720, AT3G46340, AT3G53810, AT1G78530,
AT4G08480, AT2G14750, AT1G68830, AT1G16160, AT3G58690, AT2G26830,
AT5G01950, AT3G59740, AT1G72760, AT1G30640, AT5G07280, AT5G06740,

248

Supplemental Table 4.6 (cont’d)
AT3G14370, AT1G18160, AT3G46410, AT2G41140, AT5G20480, AT2G33020,
AT1G15530, AT4G36450, AT1G76040, AT3G49060, AT4G32000, AT4G04710,
AT5G63610, AT4G13820, AT1G11280, AT3G09780, AT3G53030, AT5G65700,
AT2G30730, AT3G05660, AT1G74490, AT4G33430, AT4G18710, AT5G56460,
AT4G29180, AT3G57740, AT1G31420, AT5G55830, AT1G01740, AT1G78980,
AT5G41680, AT3G01300, AT2G41910, AT5G10290, AT3G44850, AT1G80640,
AT1G48490, AT5G39420, AT4G23220, AT2G23080, AT1G66430, AT4G00710,
AT5G65240, AT1G07560, AT3G04690, AT1G73660, AT1G51940, AT2G31010,
AT5G35370, AT1G51860, AT1G30270, AT2G26290, AT2G37710, AT1G48480,
AT5G58730, AT2G28250, AT3G08720, AT5G50180, AT4G11330, AT3G21340,
AT3G52530, AT3G50530, AT5G44290, AT2G20300, AT4G35600, AT1G21250,
AT1G53165, AT2G20470, AT4G35230, AT1G67890, AT2G47060, AT5G42120,
AT5G51350, AT1G50700, AT1G69790, AT3G49370, AT5G26150, AT3G45390,
AT4G27600, AT5G58540, AT2G18890, AT1G08590, AT5G11410, AT4G32300,
AT3G46350, AT3G54090, AT5G43320, AT1G53700, AT4G23130, AT1G10620,
AT5G40380, AT2G36350, AT5G22840, AT4G08800, AT1G61400, AT1G11300,
AT5G42440, AT3G59700, AT2G07180, AT5G20050, AT3G16030, AT1G67520,
AT5G38250, AT3G06640, AT2G29250, AT1G02970, AT5G61570, AT5G35380,
AT1G04440, AT1G61590, AT4G32250, AT5G51770, AT4G14480, AT1G73080,
AT5G60900, AT3G61960, AT2G44830, AT5G59680, AT4G23300, AT3G57830,
AT1G26970, AT2G19410, AT2G31800, AT1G61440, AT5G37790, AT5G15080,
AT5G45810, AT1G51170, AT2G19400, AT1G72540, AT5G50000, AT4G11890,
AT1G16150, AT3G47580, AT5G41990, AT4G04510, AT4G28490, AT5G59700,
AT4G02420, AT5G59660, AT5G65600, AT1G67840, AT3G57700, AT3G45420,
AT2G32680, AT5G45780, AT3G45240, AT4G34440, AT2G24370, AT1G61950,
AT1G65790, AT5G64960, AT5G61560, AT3G10660, AT4G23140, AT3G56760,
AT3G01840, AT2G23200, AT4G00720, AT3G03900, AT3G08760, AT1G51820,
AT1G48210, AT1G70530, AT5G63370, AT4G09570, AT1G59580, AT5G53320,
AT5G25440, AT4G00330, AT5G44100, AT2G30740, AT4G23180, AT2G43690,
AT5G38280, AT2G26980, AT1G54820, AT5G01540, AT1G74740, AT5G03320,
AT3G18750, AT3G27580, AT1G52540, AT3G52890, AT3G20410, AT5G05160,
AT1G70450, AT3G03940, AT1G53440, AT1G66460, AT5G63650, AT1G66980,
AT1G49100, AT3G19300, AT5G01550, AT5G35580, AT1G51910, AT1G73690,
AT5G09890, AT5G55090, AT4G27300, AT2G30040, AT1G51850, AT3G47090,
AT1G61360, AT1G71830, AT2G28590, AT4G29810, AT1G10210, AT5G20930,
AT3G61080, AT4G05200, AT5G12180, AT3G21220, AT1G76540, AT1G11340,
AT2G17520, AT1G49580, AT4G22130, AT5G58380, AT5G23170, AT1G17910,
AT4G14350, AT5G35410, AT5G60270, AT2G31500, AT1G57700, AT3G57530,
AT1G34300, AT4G28980, AT5G39020, AT5G56890, AT3G57120, AT4G29450,
AT1G17750, AT5G01920, AT4G21410, AT4G17660, AT5G57630, AT1G66910,
AT5G58940, AT3G57770, AT2G11520, AT5G24010, AT1G69220, AT1G79640,
AT1G54510, AT5G23580, AT3G59750, AT5G60090, AT1G05700, AT1G61390,
AT4G03230, AT1G06700, AT2G48010, AT1G72710, AT1G67510, AT1G19390,
AT4G02010, AT4G26510, AT2G25090, AT1G06730, AT5G27060, AT5G27790,
AT5G56790, AT3G26700, AT4G25390, AT1G19090, AT2G05060, AT1G23700,
AT3G46760, AT4G39940, AT3G53930, AT5G49660, AT5G40870, AT1G60940,
AT5G66790, AT5G45820, AT5G35390, AT5G43020, AT4G23280, AT3G57720,
AT5G10270, AT5G03730, AT4G04540, AT1G69910, AT3G56050, AT1G30570,
AT5G07180, AT3G07070, AT2G23770, AT3G45330, AT2G18170, AT2G26700,
AT5G18700, AT2G41920, AT1G79250, AT5G40030, AT1G64210, AT2G19210,
AT4G04740, AT5G12000, AT3G46140, AT3G51850, AT1G51620, AT1G69730,
AT4G25160, AT1G56720, AT1G07550, AT4G23210, AT5G03140, AT4G32710,
AT5G48740, AT4G23190, AT3G23310, AT4G22940, AT1G70110, AT2G17290,

249

Supplemental Table 4.6 (cont’d)

lipid binding
(GO:0008289)

oxidoreductase
activity
(GO:001649)

AT1G61490, AT2G39660, AT3G57730, AT1G11050, AT5G49780, AT3G63260,
AT5G46570, AT1G70130, AT4G15530, AT1G45160, AT5G37850, AT1G29740,
AT3G47570, AT1G03030, AT4G21380, AT1G50610, AT1G71410, AT4G31250,
AT5G65710, AT3G08730, AT1G08650, AT4G24400, AT4G04720, AT4G30960,
AT1G12460, AT2G17090, AT5G10530, AT3G42880, AT3G11010, AT5G55910,
AT4G26100, AT5G57015, AT1G76370, AT1G55200, AT4G28670, AT5G01020,
AT3G04530, AT5G67280, AT1G33770, AT5G03640, AT5G12090, AT5G58050,
AT4G08500, AT4G39400, AT3G50310, AT5G16500, AT3G28890, AT2G25220,
AT5G38240, AT3G24790, AT2G31880, AT5G07070, AT5G47750, AT2G35890,
AT3G06030, AT5G53450, AT5G01850, AT1G06030, AT1G76360, AT1G51880,
AT5G06940, AT4G26070, AT4G35500, AT5G16590, AT3G50000, AT5G03300,
AT5G38260, AT3G46290, AT1G07650
AT5G07540, AT4G33355, AT2G15050, AT4G08670, AT1G43665, AT1G12100,
AT1G62790, AT5G38170, AT4G27140, AT2G15325, AT1G55260, AT5G46900,
AT2G27130, AT3G61050, AT3G18280, AT2G37870, AT5G07530, AT5G55460,
AT1G32280, AT5G45560, AT3G22620, AT5G38160, AT5G55410, AT3G22580,
AT2G48130, AT4G22630, AT4G27150, AT2G13820, AT4G12470, AT3G08770,
AT1G73560, AT5G07520, AT5G56480, AT1G62500, AT5G48490, AT1G48750,
AT3G57310, AT3G20270, AT4G30880, AT5G38195, AT4G22520, AT5G62080,
AT3G22600, AT5G48485, AT1G62510, AT1G66850, AT4G22610, AT5G01870,
AT5G07550, AT3G58550, AT1G04970, AT4G12490, AT5G38180, AT3G43720,
AT5G13900, AT5G07230, AT2G10940, AT4G08530, AT3G52130, AT3G53980,
AT3G22570, AT2G48140, AT3G51590, AT1G36150, AT4G12480, AT1G12090,
AT4G12510, AT5G07510, AT4G22460, AT2G45180, AT5G46890, AT4G00165,
AT4G12550, AT1G73780, AT5G54740, AT5G53470, AT1G73890, AT5G09370,
AT4G22470, AT4G12520, AT3G22120, AT2G18370, AT5G07560, AT4G33550,
AT2G44290, AT5G52160, AT4G19040, AT4G14815, AT4G27160, AT3G07450,
AT1G73550, AT4G12530, AT4G12360, AT4G15160, AT4G22490, AT2G14846,
AT5G05960, AT5G60690, AT5G64080, AT4G12500, AT4G27170, AT1G18280,
AT5G55450
AT3G03910, AT3G61580, AT3G05260, AT1G06350, AT3G61220, AT2G46210,
AT1G14520, AT2G31360, AT5G04070, AT5G21482, AT3G03350, AT1G63380,
AT2G29330, AT5G50600, AT1G20020, AT3G49620, AT5G49740, AT1G06100,
AT1G49670, AT3G03100, AT5G06060, AT1G07440, AT4G10020, AT1G76150,
AT3G20790, AT4G09670, AT1G72190, AT5G50690, AT1G06360, AT5G59540,
AT3G03980, AT1G15140, AT5G49730, AT3G55290, AT4G20760, AT5G04900,
AT3G49630, AT3G02280, AT3G06810, AT5G18210, AT2G29340, AT4G03140,
AT1G07450, AT2G17845, AT5G07440, AT5G19200, AT1G51720, AT2G29290,
AT3G01980, AT1G67730, AT4G24050, AT5G18170, AT1G03630, AT4G13250,
AT3G08970, AT3G26760, AT3G50560, AT3G03330, AT5G50130, AT2G29350,
AT2G47140, AT2G05990, AT2G24190, AT3G50210, AT2G22260, AT3G15850,
AT1G75200, AT5G28310, AT1G52340, AT3G26770, AT5G02540, AT5G51030,
AT5G50770, AT2G23096, AT4G23420, AT1G25460, AT2G29360, AT3G15870,
AT5G10050, AT1G06090, AT1G15220, AT3G04000, AT1G12550, AT3G06060,
AT3G21420, AT5G53090, AT2G29170, AT1G03990, AT3G60370, AT5G50590,
AT4G23340, AT5G63290, AT4G26965, AT2G29370, AT1G62610, AT5G11330,
AT5G60020, AT4G04930, AT4G17370, AT1G06080, AT3G46170, AT1G34200,
AT2G47150, AT5G48440, AT4G05390, AT2G37540, AT5G56470, AT3G55310,
AT5G61830, AT1G24470, AT3G47360, AT1G64590, AT3G51680, AT5G54190,
AT1G79870, AT2G29300, AT1G57770, AT1G01800, AT2G29260, AT4G15093,
AT1G68540, AT4G11410, AT5G64250, AT3G51840, AT5G67290, AT1G54870,
AT1G10310, AT2G29150, AT2G47120, AT1G32480, AT4G05530, AT3G47350,
AT4G23430, AT5G50700, AT2G07718, AT3G42960, AT1G58300, AT5G15940,

250

Supplemental Table 4.6 (cont’d)

oxygen binding
(GO:0019825)

protein binding
(GO:0005515)

AT1G61720, AT1G06120, AT2G29310, AT2G38080, AT1G30510, AT4G09750,
AT5G65205, AT2G30670, AT3G29250, AT5G53100, AT2G47130, AT5G66190,
AT2G07727, AT3G29260, AT4G13180, AT2G29320, AT3G59710, AT4G16765,
AT4G27440, AT3G56840, AT5G50160, AT4G27760, AT1G52810, AT5G60340
AT4G31500, AT5G04660, AT3G48310, AT2G46660, AT4G39950, AT1G13110,
AT3G20960, AT3G20140, AT2G30770, AT3G26180, AT4G15110, AT3G53280,
AT2G34500, AT2G32440, AT1G55940, AT1G69500, AT4G37400, AT5G25130,
AT4G37310, AT4G31950, AT1G74110, AT5G06905, AT2G12190, AT1G64940,
AT3G14650, AT3G10570, AT2G29090, AT4G32170, AT5G51900, AT1G67110,
AT1G33720, AT4G27710, AT1G28430, AT1G57750, AT3G20100, AT1G11680,
AT3G26190, AT1G73340, AT3G56630, AT4G15350, AT4G19230, AT3G20090,
AT1G12740, AT4G31940, AT4G37410, AT3G26290, AT1G64950, AT5G09970,
AT5G25120, AT4G15360, AT2G44890, AT3G48520, AT4G37360, AT2G21910,
AT2G14100, AT3G10560, AT3G14660, AT1G13090, AT1G17060, AT3G14620,
AT4G12310, AT3G48300, AT2G30750, AT2G24180, AT3G53300, AT3G26280,
AT5G23190, AT5G05690, AT3G20080, AT3G26150, AT1G01280, AT1G13080,
AT1G01600, AT3G01900, AT1G58260, AT3G26200, AT5G44620, AT5G24900,
AT2G23220, AT3G28740, AT3G26160, AT2G45550, AT5G36110, AT2G27010,
AT5G24960, AT1G11610, AT5G14400, AT2G28860, AT5G25180, AT1G50520,
AT5G10600, AT5G57220, AT4G15330, AT1G34540, AT2G27690, AT5G48000,
AT2G46950, AT2G26170, AT5G08250, AT3G14680, AT3G26830, AT1G47620,
AT1G75130, AT5G38450, AT5G47990, AT1G64930, AT2G27000, AT1G62580,
AT5G24910, AT3G26270, AT5G67310, AT1G33730, AT1G74550, AT3G26170,
AT3G26210, AT3G13730, AT1G01190, AT5G61320, AT2G42850, AT1G11600,
AT4G39510, AT5G35715, AT3G19270, AT5G10610, AT2G28850, AT3G25180,
AT1G16400, AT2G22330, AT5G04330, AT5G42580, AT5G38970, AT3G53130,
AT5G45340, AT3G14690, AT3G48290, AT2G25160, AT1G64900, AT5G42650,
AT1G13710, AT2G45570, AT3G53305, AT2G05180, AT3G14610, AT3G52970,
AT3G26300, AT1G74540, AT5G04630, AT1G13140, AT4G13290, AT2G26710,
AT1G65670, AT5G42590, AT2G45970, AT3G20110, AT1G24540, AT3G30290,
AT5G07990, AT5G06900, AT3G26220, AT3G03470, AT4G13310, AT4G13770,
AT5G52400, AT1G31800, AT4G37340, AT1G05160, AT4G37370, AT5G63450,
AT1G66540, AT3G26310, AT4G12330, AT2G34490, AT1G13150, AT3G20120,
AT4G15300, AT2G45580, AT3G20940, AT5G05260, AT2G46960, AT4G12320,
AT4G15380, AT4G20240, AT2G16060, AT4G12300, AT4G22690, AT4G36380,
AT3G44970, AT5G02900, AT4G00360, AT4G31970, AT3G48320, AT3G14630,
AT3G61880, AT3G26125, AT1G19630, AT2G45510, AT4G39500, AT2G23180,
AT1G65340, AT3G26320, AT1G13100, AT5G52320, AT3G20130, AT5G36130,
AT1G50560, AT4G37330, AT3G20950, AT5G57260, AT5G25900, AT2G42250,
AT3G14640, AT4G39480, AT3G53290, AT5G25140, AT4G22710, AT5G58860,
AT3G44250, AT3G48280, AT1G63710, AT1G78490, AT4G37430, AT3G26330,
AT5G36220, AT3G61040, AT3G30180, AT4G37320, AT5G24950, AT3G26230,
AT1G79370, AT2G02580, AT3G48270, AT2G23190
AT1G06190, AT3G21860, AT3G21865, AT1G76490, AT3G09770, AT2G43010,
AT4G37150, AT1G23420, AT2G46280, AT1G16280, AT5G59710, AT5G33280,
AT3G23820, AT4G02020, AT2G46260, AT5G13180, AT3G10670, AT3G01280,
AT3G15150, AT3G56710, AT3G58040, AT1G50430, AT5G08130, AT5G65460,
AT1G28520, AT3G05870, AT5G06950, AT5G45680, AT5G61960, AT5G51700,
AT5G51120, AT5G16000, AT3G12690, AT3G62440, AT5G58440, AT5G27320,
AT3G53120, AT4G35620, AT4G01026, AT5G23820, AT1G78080, AT5G37500,
AT2G42830, AT5G04920, AT3G18730, AT2G32710, AT5G56860, AT1G15220,
AT5G22330, AT5G15840, AT1G70700, AT5G57900, AT2G39090, AT1G10270,
AT4G10920, AT2G25850, AT2G18840, AT2G38250, AT4G34000, AT3G28910,

251

Supplemental Table 4.6 (cont’d)
AT3G11820, AT2G34150, AT2G45980, AT5G49500, AT5G46330, AT4G00150,
AT1G15910, AT4G35090, AT5G23310, AT2G41310, AT1G56650, AT1G25490,
AT2G20080, AT3G11220, AT4G00570, AT4G20260, AT5G63880, AT1G02280,
AT1G33410, AT2G44740, AT1G48270, AT2G17950, AT5G24520, AT5G43350,
AT4G32650, AT4G19640, AT5G17690, AT3G24620, AT5G25780, AT3G50060,
AT5G20570, AT4G26090, AT5G06140, AT2G47450, AT1G19120, AT5G35750,
AT2G18710, AT2G27040, AT4G26160, AT5G40480, AT1G62360, AT1G28490,
AT3G21870, AT5G64050, AT4G25320, AT5G41990, AT1G48500, AT5G40460,
AT5G25760, AT3G14010, AT3G15540, AT4G27780, AT5G48820, AT5G43170,
AT1G30135, AT3G06560, AT4G28270, AT2G46070, AT3G53260, AT5G13190,
AT4G13250, AT1G71130, AT1G07270, AT3G60010, AT1G54320, AT3G56650,
AT1G09530, AT5G16490, AT2G07560, AT1G31350, AT1G74890, AT5G44740,
AT4G38130, AT1G30950, AT1G75410, AT1G69840, AT2G45770, AT4G28450,
AT5G04930, AT4G30260, AT4G14700, AT1G24280, AT2G32720, AT3G48160,
AT3G25500, AT4G14560, AT4G25530, AT5G03530, AT4G18290, AT1G74380,
AT2G22810, AT1G74470, AT1G15100, AT5G37600, AT4G14147, AT4G20780,
AT3G19290, AT2G14120, AT2G45190, AT5G46520, ATCG01130, AT1G73150,
AT3G18980, AT3G45640, AT5G19400, AT5G67510, AT2G31380, AT2G20160,
AT5G66570, AT5G08470, AT4G00360, AT3G45240, AT4G31730, AT5G14250,
AT4G11880, AT3G16320, AT3G16857, AT1G60430, AT4G20870, AT1G17880,
AT5G24400, AT1G01140, AT4G14960, AT1G13740, AT1G18040, AT3G19590,
AT1G08720, AT2G40030, AT3G54170, AT4G11260, AT1G17080, AT5G56290,
AT5G01410, AT2G19830, AT4G23650, AT5G01380, AT3G50070, AT2G22490,
AT4G03190, AT5G24330, AT5G14920, AT1G02140, AT1G55310, AT5G64330,
AT1G28480, AT2G29680, AT5G61010, AT1G77920, AT5G12900, AT4G17615,
AT5G17020, AT4G28840, AT3G15000, AT1G06040, AT5G06150, AT5G21940,
AT5G07280, AT5G48670, AT4G33270, AT5G55230, AT4G22920, AT2G24790,
AT4G37770, AT5G61380, AT5G02470, AT3G06400, AT5G65800, AT5G20320,
AT5G48250, AT4G26200, AT4G08980, AT3G10525, AT1G10470, AT4G38580,
AT1G16330, AT2G26650, AT1G30490, AT1G29260, AT5G51100, AT3G17880,
AT2G20890, AT3G06190, AT5G53160, AT5G43900, AT2G45740, AT3G14990,
AT3G25810, AT4G35600, AT4G29910, AT4G14713, AT2G29570, AT3G61060,
AT4G04780, AT2G38470, AT4G23810, AT5G16510, AT5G03520, AT4G39980,
AT4G25540, AT2G44900, AT5G36250, AT1G13120, AT4G35040, AT5G37055,
AT5G22220, AT4G14150, AT1G02970, AT4G14880, AT3G03450, AT5G51230,
AT4G18130, AT2G36490, AT2G26150, AT3G02470, AT5G49450, AT2G17750,
AT3G12360, AT1G15750, AT1G73830, AT5G10300, AT2G41680, AT3G63130,
AT2G37630, AT5G42190, AT4G18620, AT5G42750, AT4G26840, AT4G24940,
AT5G40930, AT1G19050, AT4G14550, AT5G63370, AT4G09570, AT1G47870,
AT3G56980, AT1G17980, AT1G55520, AT3G04680, AT3G12810, AT4G19660,
AT1G32230, AT5G05340, AT5G02110, AT4G25230, AT3G60250, AT5G27620,
AT2G26300, AT5G51450, AT2G36960, AT2G40550, AT2G40000, AT5G05760,
AT5G40280, AT4G09820, AT4G28980, AT4G27630, AT5G64070, AT3G54610,
AT3G46710, AT3G27080, AT1G16890, AT5G55000, AT3G10730, AT4G26455,
AT4G12720, AT1G53310, AT1G47220, AT3G57230, AT4G24540, AT4G17060,
AT3G18524, AT2G37040, AT4G22910, AT3G21850, AT3G05420, AT2G30250,
AT4G13520, AT4G13830, AT5G03730, AT1G64350, AT1G21700, AT3G50820,
AT5G13530, AT1G30970, AT4G25160, AT1G20140, AT5G13840, AT1G51950,
AT2G17290, AT4G21100, AT3G52750, AT3G52190, AT5G55910, AT2G33610,
AT5G51200, AT1G24260, AT5G03280, AT1G50460, AT3G61070, AT3G19720,
AT3G16050, AT2G31270, AT4G01090, AT5G16850, AT2G42880, AT5G61650,
AT5G16690, AT4G25550, AT5G45250, AT2G36910, AT4G35050, AT1G45249,
AT2G42400, AT5G22780, AT3G62410, AT2G31470, AT3G13300, AT3G45620,

252

Supplemental Table 4.6 (cont’d)
AT4G24972, AT1G32330, AT2G01120, AT5G41410, AT5G56580, AT4G11280,
AT2G22540, AT2G38880, AT1G59610, AT5G51590, AT1G18080, AT5G25350,
AT3G07780, AT5G47670, AT2G45820, AT2G27250, AT1G78600, AT5G56030,
AT1G62170, AT5G41700, AT1G18400, AT1G22770, AT1G17200, ATMG00290,
AT1G32060, AT5G10440, AT5G47010, AT4G30935, AT4G15630, AT5G41070,
AT5G17790, AT1G17840, AT1G49620, AT1G69670, AT2G40010, AT1G31140,
AT3G57130, AT3G54620, AT5G47700, AT5G35410, AT3G49850, AT2G46790,
AT1G49950, AT1G49480, AT4G20940, AT1G31360, AT3G53230, AT4G02570,
AT2G46225, AT2G43160, AT5G13300, AT1G14850, AT3G01770, AT3G15970,
AT2G27100, AT3G50630, AT4G02150, AT5G44280, AT5G49910, AT4G27950,
AT3G10540, AT5G48720, AT5G02100, AT3G57040, AT5G20850, AT3G20330,
AT2G37180, AT5G02030, AT1G75080, AT5G03340, AT3G56400, AT3G52180,
AT5G61850, AT1G68050, AT1G16610, AT4G37930, AT5G13160, AT5G52830,
AT1G80840, AT4G10090, AT3G17510, AT4G25560, AT2G44920, AT1G20930,
AT1G30210, AT3G53760, AT1G52740, AT1G10230, AT1G72450, AT1G04250,
AT5G22570, AT1G01370, AT3G17609, AT5G08330, AT2G24120, AT3G52560,
AT5G14750, AT3G48150, AT5G11300, AT1G48050, AT2G18020, AT4G32180,
AT2G16770, AT3G22590, AT5G18580, AT4G24210, AT5G01840, AT1G19850,
AT2G28060, AT4G09550, AT3G11730, AT5G63350, AT4G18700, AT5G24290,
AT1G26670, AT3G22680, AT4G22200, AT5G60550, AT5G56280, AT3G51260,
AT5G67250, AT3G03490, AT2G42810, AT4G08150, AT1G22985, AT1G66840,
AT3G11130, AT2G17820, AT3G51860, AT1G03445, AT5G10450, AT4G17460,
AT4G16110, AT3G23150, AT5G46860, AT5G09810, AT1G55610, AT1G01620,
AT5G01590, AT1G35670, AT5G35620, AT4G23750, AT1G03840, AT1G64750,
AT3G62980, AT5G13220, AT1G03190, AT2G31900, AT1G35580, AT5G39760,
AT2G36010, AT4G27500, AT1G64990, AT2G46410, AT4G02560, AT5G65700,
AT3G21630, AT4G24560, AT5G06850, AT5G44635, AT5G01820, AT3G01090,
AT1G67710, AT4G16890, AT1G09570, AT5G05410, AT1G51510, AT5G13480,
AT1G23900, AT4G04910, AT4G05420, AT1G50640, AT5G42970, AT1G65480,
AT1G11400, AT1G31770, AT4G13180, AT4G25100, AT4G37630, AT3G46510,
AT1G50240, AT4G13340, AT5G07070, AT1G64280, AT5G13790, AT5G62430,
AT5G13820, AT1G58290, AT3G52770, AT1G65620, AT2G43790, AT5G45130,
AT2G35110, AT3G01435, AT5G07090, AT2G26990, AT4G24740, AT4G29940,
AT4G34210, AT3G10572, AT5G20900, AT5G65410, AT5G04990, AT4G25570,
AT5G57050, AT3G28180, AT4G04770, AT3G09840, AT3G53570, AT4G15900,
AT2G33560, AT5G09830, AT1G01560, AT5G27080, AT1G04240, AT1G01360,
AT5G15290, AT3G25250, AT1G73590, AT5G66730, AT2G45790, AT2G38440,
AT5G18260, AT4G11110, AT5G23430, AT5G62920, AT3G16650, AT5G11530,
AT5G27150, AT1G04400, AT3G03000, AT5G63160, AT2G41140, AT1G61010,
AT1G02170, AT3G24810, AT5G63610, AT2G26040, AT4G30820, AT1G02580,
AT4G00850, AT4G18710, AT2G18160, AT2G20580, AT5G56540, AT4G32040,
AT5G67260, AT4G27330, AT4G14830, AT3G02150, AT3G11630, AT3G23780,
AT4G04720, AT3G15354, AT2G36100, AT3G22942, AT1G09415, AT1G09140,
AT3G18165, AT5G14960, AT4G27420, AT1G77740, AT1G22190, AT4G23980,
AT4G17870, AT5G05000, AT3G26090, AT3G03950, AT5G01900, AT4G08455,
AT3G04740, AT4G12570, AT4G37130, AT3G60600, AT5G52250, AT4G12020,
AT2G19560, AT1G09700, AT2G27050, AT5G01630, AT4G00355, AT5G65420,
AT5G45860, AT4G02510, AT2G03160, AT4G02680, AT1G80080, AT4G37000,
AT5G21274, AT1G16970, AT2G04240, AT4G32850, AT5G06200, AT5G55160,
AT1G65700, AT3G21150, AT3G20310, AT2G33770, AT5G55300, AT4G02195,
AT4G34110, AT1G50250, AT3G06530, AT3G01330, AT5G61600, AT1G65380,
AT3G09920, AT3G20550, AT3G52890, AT1G20590, AT5G62390, AT1G71230,
AT3G28860, AT4G39400, AT5G43830, AT1G73690, AT1G73000, AT5G61480,

253

Supplemental Table 4.6 (cont’d)
AT4G11850, AT3G21200, AT2G18915, AT1G01820, AT4G31160, AT1G08370,
AT5G13860, AT3G62770, AT3G26060, AT3G52430, AT3G07040, AT5G63110,
AT1G32900, AT5G10270, AT1G31440, AT5G46240, AT2G18170, AT3G03300,
AT4G11920, AT2G41370, AT5G10470, AT1G79690, AT3G17860, AT3G46060,
AT5G02820, AT4G00480, AT1G27630, AT1G75840, AT5G64920, AT2G19080, MPI1,
AT4G27450, AT1G59660, AT3G47620, AT3G50310, AT3G54650, AT5G39340,
AT5G47750, AT1G48760, AT1G76080, AT3G05590, AT2G03190, AT2G33310,
AT4G26070, AT5G25380, AT3G58780, AT1G09770, AT1G63650, AT3G50000,
AT3G54010, AT2G47430, AT4G12560, AT5G45870, AT3G20780, AT4G02500,
AT2G03170, AT4G22950, AT4G33000, AT1G02450, AT2G22640, AT1G28250,
AT5G60910, AT5G51660, AT1G74950, AT5G48930, AT3G08850, AT5G43060,
AT3G15210, AT3G59760, AT5G02220, AT3G57090, AT2G06005, AT3G13550,
AT2G39940, AT4G34470, AT2G13540, AT5G08590, AT4G37650, AT1G50710,
AT5G13800, AT5G65720, AT3G01320, AT2G37560, AT1G69390, AT5G42480,
AT1G30400, AT4G12780, AT4G30200, ATMG00090, AT4G33950, AT1G27930,
AT1G07500, AT5G15160, AT1G75540, AT5G03220, AT3G12280, AT5G42080,
AT5G45050, AT2G45640, AT1G24590, AT5G22290, AT5G60120, AT5G19280,
AT5G62740, AT5G04510, AT1G04260, AT1G70660, AT3G05120, AT3G57860,
AT3G02310, AT1G53720, AT1G14320, AT2G27600, AT2G45000, AT3G16830,
AT2G18960, AT2G38310, AT4G01050, AT3G48100, AT3G28030, AT1G21410,
AT3G19760, AT3G11540, AT5G66880, AT1G14920, AT3G09440, AT5G57360,
AT3G26650, AT3G02000, AT3G49670, AT4G04020, AT2G34010, AT5G13120,
AT5G63320, AT1G32640, AT5G24800, AT3G08530, AT2G41740, AT3G63010,
AT1G05020, AT3G19210, AT2G36060, AT3G55000, AT3G55005, AT3G19180,
AT3G02170, AT1G08180, AT3G24650, AT2G22430, AT1G12980, AT3G59790,
AT1G77760, AT1G52890, AT5G27600, AT1G52380, AT5G60450, AT5G60100,
AT3G51300, AT2G47510, AT5G40160, AT3G50670, AT5G50680, AT3G23050,
AT4G16250, AT2G36270, AT5G56210, AT3G15730, AT1G17790, AT3G24495,
AT2G40340, AT3G09260, AT5G06310, AT1G29680, AT3G57260, AT4G33010,
AT5G05680, AT2G19430, AT2G36307, AT1G06110, AT5G50750, AT5G43070,
AT1G09340, AT2G02560, AT2G05520, AT3G10340, AT2G36890, AT5G20810,
AT4G34460, AT3G54840, AT5G38110, AT1G27320, AT2G34650, AT1G80350,
AT4G37460, AT5G48380, AT5G13290, AT1G07370, AT2G30330, AT1G23860,
AT1G52410, AT5G48160, AT5G45110, AT3G09900, AT3G21510, AT5G20920,
AT3G28200, AT1G07880, AT3G20770, AT4G33945, AT5G03940, AT2G38620,
AT5G58220, AT1G23260, AT2G42580, AT5G04230, AT3G52850, AT5G03540,
AT1G10760, AT3G27920, AT1G53650, AT5G59880, AT4G24660, AT5G53490,
AT3G14110, AT3G61570, AT2G38940, AT5G53470, AT3G06110, AT3G48680,
AT5G49160, AT5G28770, AT1G47230, AT5G23880, AT5G23730, AT3G19770,
AT3G11550, AT1G43850, AT2G05120, AT2G20000, AT4G23570, AT4G18040,
AT4G36810, AT5G14620, AT5G23080, AT1G47128, AT3G18910, AT2G45450,
AT3G45780, AT5G23260, AT3G22880, AT3G45140, AT4G29130, AT1G44800,
AT4G32010, AT1G68640, AT3G05040, AT2G23350, AT5G67580, AT2G40750,
AT5G66055, AT1G66390, AT1G49760, AT2G01950, AT3G07880, AT1G79830,
AT2G18790, AT1G76310, AT5G63790, AT5G21010, AT1G75820, AT5G64900,
AT1G69600, AT5G35840, AT3G43920, AT1G55490, AT5G63310, AT1G02860,
AT1G12220, AT2G22240, AT2G02950, AT5G63860, AT4G19030, AT3G60890,
AT1G09020, AT1G75950, AT3G56900, AT1G78870, AT3G43810, AT5G20240,
AT1G44110, AT2G24765, AT4G17750, AT1G17745, AT3G14470, AT4G32551,
AT5G43080, AT1G26840, AT1G31280, AT4G08480, AT4G29160, AT5G02200,
AT4G02740, AT2G46340, AT5G07120, AT4G28560, AT5G03150, AT2G24490,
AT2G01730, AT3G54850, AT2G13560, AT2G25700, AT3G60360, AT5G20480,
AT1G23490, AT3G61140, AT1G30460, AT3G62090, AT1G80170, AT5G48170,

254

Supplemental Table 4.6 (cont’d)
AT4G35230, AT5G53480, AT1G21780, AT1G54040, AT3G12400, AT5G18620,
AT2G21470, AT4G20380, AT1G73790, AT5G57160, AT3G26510, AT1G47750,
AT1G10390, AT1G24150, AT2G30490, AT2G39770, AT5G12880, AT2G28380,
AT5G16560, AT5G23720, AT3G48300, AT3G63140, AT5G14520, AT4G39710,
AT3G62720, AT2G18300, AT4G36800, AT1G17760, AT1G73080, AT2G45280,
AT2G41100, AT3G11410, AT5G11440, AT3G43440, AT2G46310, AT5G46210,
AT4G35900, AT1G77140, AT3G62030, AT4G14870, AT3G29160, AT5G57740,
AT3G48090, AT3G48360, AT4G23140, AT4G00650, AT4G20360, AT4G36480,
AT1G32530, AT2G24840, AT5G05440, AT2G41430, AT3G23100, AT2G23430,
AT5G11710, AT2G23380, AT3G25070, AT2G04660, AT5G39510, AT2G28800,
AT4G24280, AT5G67480, AT5G64350, AT3G20000, AT1G55805, AT1G28420,
AT4G19003, AT2G36250, AT3G51960, AT1G09030, AT4G26020, AT1G76540,
AT1G02090, AT2G46970, AT4G03280, AT5G40740, AT3G56970, AT5G17620,
AT4G02640, AT3G57530, AT1G63020, AT4G29510, AT5G20720, AT3G54220,
AT3G15880, AT2G22670, AT3G25980, AT1G67580, AT2G26570, AT3G62800,
AT1G50030, AT1G26830, AT4G29170, AT3G50360, AT1G53090, AT5G44200,
AT1G27300, AT2G32400, AT3G15660, AT3G05380, AT1G06770, AT3G27010,
AT5G13570, AT1G07130, AT4G27860, AT1G06950, AT4G13930, AT3G43300,
AT1G16240, AT3G06590, AT3G53430, AT2G26700, AT4G02060, AT1G79250,
AT2G46020, AT5G58590, AT1G29170, AT2G25000, AT5G50580, AT1G65290,
AT2G31570, AT3G56370, AT1G30270, AT1G32310, AT4G32980, AT4G14310,
AT4G37530, AT5G53290, AT4G09960, AT3G61600, AT5G03455, AT1G47580,
AT5G58040, AT2G06530, AT3G29575, AT4G28640, AT1G25540, AT2G33460,
AT3G53610, AT5G23860, AT2G39760, AT2G42010, AT2G17730, AT1G80680,
AT1G73500, AT5G26980, AT5G15800, AT5G27100, AT4G18020, AT2G41110,
AT2G16070, AT2G32950, AT1G04940, AT1G19180, AT2G22840, AT5G47100,
AT1G01040, AT3G19220, AT4G15510, AT4G10760, AT5G27030, AT5G23040,
AT5G41790, AT4G31580, AT2G40730, AT5G66030, ATCG00500, AT4G11140,
AT3G48590, AT1G22070, AT4G36290, AT4G16420, AT4G33690, AT4G35800,
AT4G19170, AT5G64960, AT3G55120, AT1G77080, AT3G25710, AT3G51970,
AT5G63510, AT5G48570, AT5G52220, AT3G23000, AT3G54180, AT1G54610,
AT2G37340, AT5G01370, AT5G47080, AT1G59750, AT2G01620, AT2G46700,
AT3G04000, AT3G24440, AT4G32570, AT1G12390, AT3G03740, AT5G59430,
AT1G71310, AT2G25490, AT1G35160, AT5G48400, AT1G69400, AT2G27370,
AT3G24590, AT5G48990, AT5G55190, AT5G13490, AT3G25882, AT1G75010,
AT3G15500, AT4G21560, AT5G20730, AT4G26610, AT2G31985, AT4G28910,
AT5G55990, AT5G21170, AT5G20910, AT4G05000, AT3G18780, AT4G02070,
AT1G08550, AT3G57290, AT3G13110, AT3G13445, AT1G70510, AT4G27920,
AT4G35580, AT1G71860, AT1G34030, AT5G45010, AT2G01570, AT2G38170,
AT4G29830, AT1G29010, AT1G75340, AT1G24310, AT3G14120, AT5G28640,
AT2G30360, AT5G19000, AT5G11260, AT4G39800, AT5G15580, AT3G25230,
AT4G30190, AT1G08830, AT4G35100, AT1G74500, AT5G13930, AT2G16850,
AT4G38460, AT5G04240, AT1G04310, AT1G66740, AT5G11390, ATCG00190,
AT5G16830, AT2G44950, AT5G08720, AT1G08780, AT2G30580, AT1G19350,
AT3G02280, AT5G23670, AT5G46760, AT5G24020, AT3G63500, AT4G33510,
AT1G72770, AT1G04550, AT5G27000, AT2G40890, AT1G66340, AT5G40810,
AT5G10350, AT3G23670, AT5G24110, AT4G01900, AT5G09250, AT5G60410,
AT3G33520, AT1G56330, AT5G24590, AT3G61190, AT2G26280, AT2G15400,
AT1G05200, AT2G40380, AT3G50500, AT2G01980, AT2G40470, AT3G07560,
AT1G77180, AT3G51630, AT5G40030, AT1G17730, AT1G02340, AT2G33270,
AT4G26750, AT2G26350, AT5G06960, AT4G29350, AT1G76920, AT4G12620,
AT4G26570, AT2G01760, AT3G57350, AT2G46600, AT3G46580, AT3G20600,
AT3G47500, AT4G37520, AT1G02840, AT1G09270, AT5G65930, AT4G19990,

255

Supplemental Table 4.6 (cont’d)

proteolysis
(GO:0006508)

AT5G20010, AT4G14180, AT5G08450, AT1G69120, AT2G34180, AT5G26860,
AT5G66750, ATCG00020, AT3G49700, AT1G47260, AT5G66280, AT5G07300,
AT1G05460, AT2G30470, AT1G01160, AT1G69690, AT5G57450, AT1G78770,
AT4G01370, AT5G18410, AT5G04140, AT1G22920, AT2G04550, AT3G43700,
AT1G65030, AT3G02230, AT1G71830, AT2G45650, AT4G08920, AT1G01910,
AT5G09800, AT3G62420, AT1G10690, AT4G14220, AT5G19330, AT3G19150,
AT1G74310, AT3G57870, AT5G09790, AT3G59220, AT5G03415, AT3G13920,
AT1G46408, AT5G20930, AT2G42260, AT2G24540, AT3G20740, AT1G70310,
AT1G10970, AT5G58230, AT1G04450, AT3G06720, AT5G12390, AT1G71800,
AT2G05380, AT2G45660, AT3G21430, AT2G04890, AT3G14080, AT1G21250,
AT5G02500, AT3G21640, AT5G22880, AT3G47430, AT2G26760, AT4G31800,
AT1G10940, AT4G02440, AT3G10650, AT5G04470, AT2G27960, AT5G16620,
AT1G15570, AT5G58550, AT2G30770, AT1G21750, AT5G44550, AT5G04870,
AT1G65650, AT1G70210, AT4G14110, AT5G57110, AT5G62000, AT5G41315,
AT5G53120, AT2G21240, AT5G54490, AT5G04900, AT3G17590, AT3G12250,
AT5G22640, AT4G09010, AT1G66750, AT4G04740, AT3G29350, AT5G42520,
AT2G45140, AT3G16770, AT1G05560, AT2G19110, AT3G48750, AT1G73360,
AT1G01510, AT1G21970, AT1G61790, AT1G53510, AT2G40940, AT1G22275,
AT2G41620, AT5G10030, AT5G17560, AT4G37490, AT4G33430, AT3G63400,
AT2G18040, AT5G09260, AT3G16420, AT2G40330, AT4G03080, AT5G40330,
AT4G30960, AT5G25890, AT1G76260, AT5G44180, AT5G20020, AT5G24270,
AT5G46790, AT5G35790, AT1G48380, AT1G32500, AT1G56250, AT1G49720,
AT3G05280, AT4G03270, AT4G17710, AT2G44610, AT5G52550, AT2G36350,
AT1G48410, AT2G46830, AT4G33650, AT1G80490, AT4G27160, AT1G17380,
AT3G04730, AT1G62300, AT3G46590, AT1G12110, AT1G01640, AT2G36990,
AT1G64860, AT4G32910, AT3G08900, AT5G55280, AT5G01640, AT3G57560,
AT5G65210, AT5G54640, AT5G64630, AT5G22770, AT3G13460, AT1G03000,
AT5G62810, AT5G02420, AT5G08120, AT2G27970, AT5G16320, AT4G13870,
AT4G37940, AT5G59380, AT5G44560, AT2G38280, AT3G18690, AT4G39890,
AT2G29100, AT1G74740, AT1G20610, AT4G29810, AT4G34160, AT3G06380,
AT5G13060, AT5G15210, AT3G61630, AT3G44530, AT5G02490, AT1G30330,
AT4G04885, AT2G05210, AT2G29210, AT1G79040, AT1G20780, AT4G14720,
AT3G55840, AT1G79280, AT2G31500, AT5G14270, AT5G57380, AT5G15850,
AT3G59380, AT1G10060, AT5G49060, AT3G54820, AT1G24510, AT1G02980,
AT5G18400, AT4G36930, AT2G18290, AT4G23450, AT4G00180, AT5G67570,
AT2G20180, AT4G31710, AT3G11520, AT5G14070, AT2G28160, AT5G50950,
AT1G14740, AT3G23380, AT4G15410, AT1G26110, AT2G40670, AT1G80670,
AT1G43700, AT4G15090, AT1G32400, AT3G51920, AT3G08730, AT2G01830,
AT3G23030, AT4G30840, AT5G47120, AT1G09070, AT2G35720, AT5G52120,
AT2G32980, AT1G56260, AT4G08500, AT4G26080, AT1G12360, AT3G54710,
AT2G36160, AT5G25220, AT4G17490, AT3G50410, AT4G16144, AT5G39660,
AT5G61210, AT5G60340
AT4G31500, AT5G04660, AT3G48310, AT2G46660, AT4G39950, AT1G13110,
AT3G20960, AT3G20140, AT2G30770, AT3G26180, AT4G15110, AT3G53280,
AT2G34500, AT2G32440, AT1G55940, AT1G69500, AT4G37400, AT5G25130,
AT4G37310, AT4G31950, AT1G74110, AT5G06905, AT2G12190, AT1G64940,
AT3G14650, AT3G10570, AT2G29090, AT4G32170, AT5G51900, AT1G67110,
AT1G33720, AT4G27710, AT1G28430, AT1G57750, AT3G20100, AT1G11680,
AT3G26190, AT1G73340, AT3G56630, AT4G15350, AT4G19230, AT3G20090,
AT1G12740, AT4G31940, AT4G37410, AT3G26290, AT1G64950, AT5G09970,
AT5G25120, AT4G15360, AT2G44890, AT3G48520, AT4G37360, AT2G21910,
AT2G14100, AT3G10560, AT3G14660, AT1G13090, AT1G17060, AT3G14620,
AT4G12310, AT3G48300, AT2G30750, AT2G24180, AT3G53300, AT3G26280,

256

Supplemental Table 4.6 (cont’d)

response to auxin
(GO:0009733)

AT5G23190, AT5G05690, AT3G20080, AT3G26150, AT1G01280, AT1G13080,
AT1G01600, AT3G01900, AT1G58260, AT3G26200, AT5G44620, AT5G24900,
AT2G23220, AT3G28740, AT3G26160, AT2G45550, AT5G36110, AT2G27010,
AT5G24960, AT1G11610, AT5G14400, AT2G28860, AT5G25180, AT1G50520,
AT5G10600, AT5G57220, AT4G15330, AT1G34540, AT2G27690, AT5G48000,
AT2G46950, AT2G26170, AT5G08250, AT3G14680, AT3G26830, AT1G47620,
AT1G75130, AT5G38450, AT5G47990, AT1G64930, AT2G27000, AT1G62580,
AT5G24910, AT3G26270, AT5G67310, AT1G33730, AT1G74550, AT3G26170,
AT3G26210, AT3G13730, AT1G01190, AT5G61320, AT2G42850, AT1G11600,
AT4G39510, AT5G35715, AT3G19270, AT5G10610, AT2G28850, AT3G25180,
AT1G16400, AT2G22330, AT5G04330, AT5G42580, AT5G38970, AT3G53130,
AT5G45340, AT3G14690, AT3G48290, AT2G25160, AT1G64900, AT5G42650,
AT1G13710, AT2G45570, AT3G53305, AT2G05180, AT3G14610, AT3G52970,
AT3G26300, AT1G74540, AT5G04630, AT1G13140, AT4G13290, AT2G26710,
AT1G65670, AT5G42590, AT2G45970, AT3G20110, AT1G24540, AT3G30290,
AT5G07990, AT5G06900, AT3G26220, AT3G03470, AT4G13310, AT4G13770,
AT5G52400, AT1G31800, AT4G37340, AT1G05160, AT4G37370, AT5G63450,
AT1G66540, AT3G26310, AT4G12330, AT2G34490, AT1G13150, AT3G20120,
AT4G15300, AT2G45580, AT3G20940, AT5G05260, AT2G46960, AT4G12320,
AT4G15380, AT4G20240, AT2G16060, AT4G12300, AT4G22690, AT4G36380,
AT3G44970, AT5G02900, AT4G00360, AT4G31970, AT3G48320, AT3G14630,
AT3G61880, AT3G26125, AT1G19630, AT2G45510, AT4G39500, AT2G23180,
AT1G65340, AT3G26320, AT1G13100, AT5G52320, AT3G20130, AT5G36130,
AT1G50560, AT4G37330, AT3G20950, AT5G57260, AT5G25900, AT2G42250,
AT3G14640, AT4G39480, AT3G53290, AT5G25140, AT4G22710, AT5G58860,
AT3G44250, AT3G48280, AT1G63710, AT1G78490, AT4G37430, AT3G26330,
AT5G36220, AT3G61040, AT3G30180, AT4G37320, AT5G24950, AT3G26230,
AT1G79370, AT2G02580, AT3G48270, AT2G23190
AT5G06300, AT1G80680, AT4G37390, AT3G43120, AT5G13370, AT1G74950,
AT4G13790, AT2G46690, AT1G19220, AT1G72430, AT2G35940, AT1G19180,
AT5G47370, AT1G64520, AT3G06490, AT1G78100, AT2G26740, AT3G12955,
AT5G18060, AT3G07390, AT5G27420, AT3G52400, AT5G13300, AT5G59780,
AT4G11280, AT5G03310, AT2G24850, AT3G07370, AT4G16420, AT3G28210,
AT1G19640, AT2G38120, AT5G37020, AT2G24400, AT3G55120, AT1G59500,
AT1G48660, CPR6, AT2G47460, AT1G29510, AT4G34390, AT2G27690, HCA,
AT2G21220, AT2G33860, AT5G01270, AT3G09980, AT5G09810, AT4G34710,
AT1G59750, AT3G24280, AT4G14550, AT2G25790, AT2G44840, AT2G47260,
AT1G29460, AT2G28085, SAR1, AT5G50760, AT5G59430, AT3G49850, AT1G49950,
AT1G15520, AT3G28910, AT4G01370, AT5G59220, AT2G39370, AT5G27780,
AT5G18020, AT5G20820, AT4G02570, AT4G32810, AT3G25880, AT1G56150,
AT1G22920, AT3G14050, AT2G04550, AT5G65940, AT3G15500, AT3G11820,
AT1G09540, AT5G20730, AT1G80390, AT5G54500, AT3G01220, AT3G63010,
AT5G57560, AT1G74840, AT2G06050, AT4G00080, AT1G56650, AT1G25490,
AT5G13930, AT2G45210, AT3G24650, AT1G33410, AT3G09940, AT5G24520,
AT2G01570, AT1G52890, AT4G16780, AT5G06960, AT2G46830, AT5G20570,
AT1G29420, AT5G05730, AT3G03850, AT3G23050, AT5G02840, AT1G69270,
AT5G01490, AT1G75580, AT5G39610, ICR2, AT5G53590, AT4G34760, AT2G31180,
AT1G16510, AT2G25930, AT1G04250, AT3G17600, AT4G18010, AT3G61900,
AT4G21440, AT3G04730, AT3G15540, AT5G07700, AT1G04100, AT2G02560,
AT2G21210, AT1G64060, AT2G46070, AT5G20810, AT1G04550, AT2G46270,
AT4G25030, AT3G26760, AT2G34650, AT1G75590, AT3G11260, AT1G66340,
AT3G58190, AT1G19850, AT4G38630, AT2G19690, AT2G16580, AT4G32690,
AT2G04160, AT5G65670, AT4G34790, AT3G12830, AT4G15430, AT3G20770,

257

Supplemental Table 4.6 (cont’d)

response to chitin
(GO:0010200)

AT4G33940, AT1G28480, AT4G34810, AT1G18890, AT1G29500, AT1G52830,
AT2G23170, AT1G76190, AT3G09870, AT1G34670, AT5G51470, AT5G12330,
AT4G12410, AT3G20830, AT5G37260, AT1G15580, AT1G49010, AT3G47600,
AT4G14560, AT3G62980, AT5G13220, AT1G23160, AT4G38850, AT5G57090,
AT4G00880, AT1G54060, AT4G01280, AT2G34600, AT1G54100, AT2G22810,
AT4G18950, AT3G22650, AT3G16350, AT1G16540, AT5G18030, AT4G03400,
AT1G28370, AT2G20000, AT2G35270, AT4G16890, AT1G47510, AT4G31320,
AT4G34780, AT3G48360, AT1G61120, AT3G19580, AT1G14000, AT4G26400,
AT2G18010, AT2G41820, AT2G34720, AT1G43040, AT5G67580, AT2G36910,
AT2G43790, AT5G56290, AT4G05100, AT4G22620, AT4G29940, AT1G29430,
AT2G22240, AT1G70000, AT3G03820, AT4G03190, AT4G38860, AT5G42410,
AT5G54490, AGE2, AT4G34770, AT1G74100, AT2G46510, AT1G73730, AT5G40770,
AT5G55120, AT1G53230, AT1G01560, AT1G04240, AT1G29440, AT5G13350,
AT4G17615, AT5G51640, AT1G28130, AT1G73590, AT3G05630, AT2G06850,
AT1G17345, AT1G15430, AT3G53250, AT2G37030, AT1G20470, AT5G63160,
AT2G39550, AT1G19840, AT3G21110, AT4G18710, AT1G69260, AT2G26710,
AT5G67300, AT5G57420, AT4G34800, AT4G23915, AT5G08640, AT1G77690,
AT2G21200, AT5G07990, AT1G24590, AT5G25890, AT4G30080, AT1G27740,
AT2G25170, AT5G14960, AT5G19140, AT5G64890, AT1G74660, AT3G26090,
AT5G15310, GUP1, GUP2, AT4G20380, AT3G59900, AT5G66260, AT4G38840,
AT1G48410, AT2G19560, AT2G36210, AT1G09700, AT1G15050, AT5G37770,
AT1G17380, AT5G22220, AT3G60690, AT5G45710, AT1G19830, AT3G20220,
AT4G36800, AT5G13360, AT4G29080, AT1G29490, AT1G26870, AT1G15750,
AT1G08030, AT3G16500, AT2G37630, AT5G18050, AT5G57740, AT1G31340,
AT2G47750, AT4G33880, AT5G43700, AT1G63840, AT4G36740, AT2G05710,
AT4G27260, AT1G48670, AT1G18570, AT1G22640, AT1G71230, AT3G28860,
AT3G62100, AT4G14430, AT4G36110, AT5G17300, AT3G03830, AT1G30330,
AT1G10210, AT2G14960, AT2G01200, AT1G32230, AT4G27410, AT2G47190,
AT3G57530, AT2G22670, AT1G29450, AT5G10990, AT5G66700, AT1G79130,
AT2G46370, AT5G50120, AT5G67480, AT1G01060, AT1G54990, AT4G19690,
AT3G02260, AT5G18010, AT4G37295, AT1G17520, AT4G32280, AT4G12550,
AT4G32880, AT5G18080, AT4G13520, AT4G09530, AT1G06400, AT5G54510,
AT4G28640, AT3G51200, AT4G37610, AT3G23250, AT3G26790, AT1G57560,
AT5G13380, AT2G47000, AT3G09600, AT1G51950, AT2G17290, SSA-2, AT3G23030,
AT1G48690, AT3G63300, AT3G55730, AT1G27730, AT1G63720, AT4G39403,
AT4G26080, AT2G34680, AT2G46990, AT3G03847, AT3G03840, AT2G33310,
AT3G50410, AT5G38895, AT4G34750
AT5G06300, AT1G80680, AT4G37390, AT3G43120, AT5G13370, AT1G74950,
AT4G13790, AT2G46690, AT1G19220, AT1G72430, AT2G35940, AT1G19180,
AT5G47370, AT1G64520, AT3G06490, AT1G78100, AT2G26740, AT3G12955,
AT5G18060, AT3G07390, AT5G27420, AT3G52400, AT5G13300, AT5G59780,
AT4G11280, AT5G03310, AT2G24850, AT3G07370, AT4G16420, AT3G28210,
AT1G19640, AT2G38120, AT5G37020, AT2G24400, AT3G55120, AT1G59500,
AT1G48660, CPR6, AT2G47460, AT1G29510, AT4G34390, AT2G27690, HCA,
AT2G21220, AT2G33860, AT5G01270, AT3G09980, AT5G09810, AT4G34710,
AT1G59750, AT3G24280, AT4G14550, AT2G25790, AT2G44840, AT2G47260,
AT1G29460, AT2G28085, SAR1, AT5G50760, AT5G59430, AT3G49850, AT1G49950,
AT1G15520, AT3G28910, AT4G01370, AT5G59220, AT2G39370, AT5G27780,
AT5G18020, AT5G20820, AT4G02570, AT4G32810, AT3G25880, AT1G56150,
AT1G22920, AT3G14050, AT2G04550, AT5G65940, AT3G15500, AT3G11820,
AT1G09540, AT5G20730, AT1G80390, AT5G54500, AT3G01220, AT3G63010,
AT5G57560, AT1G74840, AT2G06050, AT4G00080, AT1G56650, AT1G25490,
AT5G13930, AT2G45210, AT3G24650, AT1G33410, AT3G09940, AT5G24520,

258

Supplemental Table 4.6 (cont’d)

RNA binding
(GO:0003723)

AT2G01570, AT1G52890, AT4G16780, AT5G06960, AT2G46830, AT5G20570,
AT1G29420, AT5G05730, AT3G03850, AT3G23050, AT5G02840, AT1G69270,
AT5G01490, AT1G75580, AT5G39610, ICR2, AT5G53590, AT4G34760, AT2G31180,
AT1G16510, AT2G25930, AT1G04250, AT3G17600, AT4G18010, AT3G61900,
AT4G21440, AT3G04730, AT3G15540, AT5G07700, AT1G04100, AT2G02560,
AT2G21210, AT1G64060, AT2G46070, AT5G20810, AT1G04550, AT2G46270,
AT4G25030, AT3G26760, AT2G34650, AT1G75590, AT3G11260, AT1G66340,
AT3G58190, AT1G19850, AT4G38630, AT2G19690, AT2G16580, AT4G32690,
AT2G04160, AT5G65670, AT4G34790, AT3G12830, AT4G15430, AT3G20770,
AT4G33940, AT1G28480, AT4G34810, AT1G18890, AT1G29500, AT1G52830,
AT2G23170, AT1G76190, AT3G09870, AT1G34670, AT5G51470, AT5G12330,
AT4G12410, AT3G20830, AT5G37260, AT1G15580, AT1G49010, AT3G47600,
AT4G14560, AT3G62980, AT5G13220, AT1G23160, AT4G38850, AT5G57090,
AT4G00880, AT1G54060, AT4G01280, AT2G34600, AT1G54100, AT2G22810,
AT4G18950, AT3G22650, AT3G16350, AT1G16540, AT5G18030, AT4G03400,
AT1G28370, AT2G20000, AT2G35270, AT4G16890, AT1G47510, AT4G31320,
AT4G34780, AT3G48360, AT1G61120, AT3G19580, AT1G14000, AT4G26400,
AT2G18010, AT2G41820, AT2G34720, AT1G43040, AT5G67580, AT2G36910,
AT2G43790, AT5G56290, AT4G05100, AT4G22620, AT4G29940, AT1G29430,
AT2G22240, AT1G70000, AT3G03820, AT4G03190, AT4G38860, AT5G42410,
AT5G54490, AGE2, AT4G34770, AT1G74100, AT2G46510, AT1G73730, AT5G40770,
AT5G55120, AT1G53230, AT1G01560, AT1G04240, AT1G29440, AT5G13350,
AT4G17615, AT5G51640, AT1G28130, AT1G73590, AT3G05630, AT2G06850,
AT1G17345, AT1G15430, AT3G53250, AT2G37030, AT1G20470, AT5G63160,
AT2G39550, AT1G19840, AT3G21110, AT4G18710, AT1G69260, AT2G26710,
AT5G67300, AT5G57420, AT4G34800, AT4G23915, AT5G08640, AT1G77690,
AT2G21200, AT5G07990, AT1G24590, AT5G25890, AT4G30080, AT1G27740,
AT2G25170, AT5G14960, AT5G19140, AT5G64890, AT1G74660, AT3G26090,
AT5G15310, GUP1, GUP2, AT4G20380, AT3G59900, AT5G66260, AT4G38840,
AT1G48410, AT2G19560, AT2G36210, AT1G09700, AT1G15050, AT5G37770,
AT1G17380, AT5G22220, AT3G60690, AT5G45710, AT1G19830, AT3G20220,
AT4G36800, AT5G13360, AT4G29080, AT1G29490, AT1G26870, AT1G15750,
AT1G08030, AT3G16500, AT2G37630, AT5G18050, AT5G57740, AT1G31340,
AT2G47750, AT4G33880, AT5G43700, AT1G63840, AT4G36740, AT2G05710,
AT4G27260, AT1G48670, AT1G18570, AT1G22640, AT1G71230, AT3G28860,
AT3G62100, AT4G14430, AT4G36110, AT5G17300, AT3G03830, AT1G30330,
AT1G10210, AT2G14960, AT2G01200, AT1G32230, AT4G27410, AT2G47190,
AT3G57530, AT2G22670, AT1G29450, AT5G10990, AT5G66700, AT1G79130,
AT2G46370, AT5G50120, AT5G67480, AT1G01060, AT1G54990, AT4G19690,
AT3G02260, AT5G18010, AT4G37295, AT1G17520, AT4G32280, AT4G12550,
AT4G32880, AT5G18080, AT4G13520, AT4G09530, AT1G06400, AT5G54510,
AT4G28640, AT3G51200, AT4G37610, AT3G23250, AT3G26790, AT1G57560,
AT5G13380, AT2G47000, AT3G09600, AT1G51950, AT2G17290, SSA-2, AT3G23030,
AT1G48690, AT3G63300, AT3G55730, AT1G27730, AT1G63720, AT4G39403,
AT4G26080, AT2G34680, AT2G46990, AT3G03847, AT3G03840, AT2G33310,
AT3G50410, AT5G38895, AT4G34750
AT3G50670, AT4G26650, AT1G49760, AT5G43110, AT1G14640, AT3G10360,
AT5G18110, AT5G55670, AT4G36690, AT4G16830, AT3G55460, AT2G31890,
AT1G67770, AT2G42890, AT4G13860, AT2G43640, AT3G20250, AT2G47220,
AT2G14870, AT2G07734, AT5G19350, AT5G18810, AT4G29090, AT1G71720,
AT4G22380, AT3G20890, AT4G34110, AT1G28090, AT1G27650, AT5G55100,
AT1G18050, AT3G55340, AT5G50250, AT1G07350, AT1G35730, AT5G19030,
AT3G11400, AT1G60000, AT2G47580, AT5G47210, AT1G64810, AT4G09040,

259

Supplemental Table 4.6 (cont’d)
AT1G74350, AT3G07250, AT2G19380, AT3G26120, AT1G74230, AT5G60170,
AT4G27000, AT5G12190, AT3G19130, AT4G27490, AT4G02430, AT3G54770,
AT3G10845, AT5G07060, AT3G07810, AT3G51950, AT1G60650, AT5G54900,
AT3G61620, AT3G63450, AT1G53120, ATCG00040, AT3G27700, AT3G02320,
AT1G33470, AT3G20550, AT1G58470, AT5G60960, AT3G56860, AT1G09340,
AT1G55310, AT1G80930, AT1G22760, AT1G50300, AT3G27750, AT2G30260,
AT4G24270, AT4G10110, AT4G26370, AT3G49430, AT1G49600, AT1G17640,
AT3G45630, AT3G49390, AT5G04210, AT3G15010, AT5G59860, AT3G26560,
AT5G61780, AT5G55550, AT3G46020, AT3G12680, AT4G17610, AT3G23830,
AT1G06960, AT5G23690, AT3G52150, AT1G34140, AT2G37510, AT2G22090,
AT1G13190, AT1G47490, AT3G56330, AT4G25500, AT1G62670, AT5G14580,
AT1G09140, AT1G01760, AT2G05160, AT4G08840, AT3G55280, AT1G09230,
AT1G22240, AT4G25630, AT2G29190, AT5G06520, AT1G13730, AT5G18180,
AT5G51120, AT3G16810, AT1G12800, AT1G14650, AT2G33410, AT2G29580,
AT5G52040, AT1G33520, AT4G03110, AT5G60110, AT2G19870, AT5G58130,
AT5G07350, AT5G08620, AT5G10350, AT5G16260, AT5G42820, AT4G12600,
AT5G09610, AT5G11530, AT3G48830, AT5G24440, AT1G29590, AT2G47310,
AT1G56030, AT3G06970, AT5G19960, AT1G07360, AT4G25880, AT3G03710,
AT2G29200, AT1G24450, AT3G19090, AT3G13570, AT2G23350, AT1G15910,
AT4G11175, AT1G43860, AT2G25900, AT1G16610, AT4G10610, AT5G53180,
AT5G53720, AT1G30460, AT3G13224, AT5G02250, AT4G15520, AT1G76460,
AT1G29550, AT1G30010, AT3G14450, AT5G44200, AT3G53460, AT3G21215,
AT2G18510, AT1G53720, AT5G46250, AT1G09150, AT1G22910, AT5G25060,
AT4G38020, AT1G53650, AT4G19610, AT2G21690, AT2G03640, AT1G60830,
AT5G48870, AT3G16380, AT1G37140, AT5G64200, AT2G33435, AT3G03920,
AT3G43920, AT3G08010, AT2G39460, AT2G21660, AT5G05720, AT2G43960,
AT5G04600, AT1G01080, AT4G16280, AT5G57870, AT4G36020, AT5G47620,
AT3G52120, AT3G29390, AT1G71800, AT5G30510, AT3G21100, AT2G29140,
AT4G17520, AT3G22310, AT2G43410, AT3G23700, AT4G15417, AT3G26420,
AT3G10400, AT1G45100, AT2G24350, AT4G16200, AT1G21620, AT1G18630,
AT1G43190, AT1G71770, AT3G61860, AT4G24770, AT4G34730, AT4G39260,
AT2G46780, AT1G78260, AT2G39120, AT4G14300, AT4G12640, AT2G33440,
AT1G60080, AT3G25150, AT5G61030, AT4G20030, AT2G39260, AT5G41690,
AT3G08000, AT1G77680, AT2G21440, AT5G66010, AT5G07290, AT2G39140,
AT1G22330, AT3G01150, AT5G26880, AT5G60180, AT1G29400, AT5G04280,
AT2G41060, AT5G52490, AT3G46210, AT5G64390, AT5G20320, AT2G22100,
AT4G18120, AT5G15390, AT5G65260, AT4G18040, AT5G43090, AT1G35750,
AT5G28390, AT4G00830, AT2G17510, AT1G69250, AT2G36660, AT5G06210,
AT1G73490, AT2G37220, AT2G43370, AT3G47120, AT3G07750, AT5G23080,
AT1G02840, AT5G10800, AT5G48650, AT5G08180, AT2G44710, AT1G51510,
AT5G15750, AT4G31200, AT2G46610, AT5G54580, AT3G12990, AT5G60980,
AT3G13700, AT1G78160, AT1G73530, AT3G04500, AT5G58040, AT5G43960,
AT3G60500, AT5G59280, AT5G15810, AT4G13850, AT4G36960, AT2G35410,
AT5G56510, AT5G06000, AT1G11650, AT1G72320, AT1G20880, AT1G35850,
AT4G32720, AT2G24050, AT5G51410, AT2G20490, AT3G12640, AT5G61960,
AT3G11964, AT1G10320, AT3G52660, AT3G20930, AT1G47500, AT5G20160,
AT4G37510, AT5G35620, AT2G17580, AT1G32790, AT2G40290, AT5G40490,
AT5G47320, AT3G52380, AT1G60900, AT5G46920, AT5G03580, AT5G53680,
AT5G05470, AT5G46840, AT1G03457

260

Supplemental Table 4.6 (cont’d)
transferase
activity,
transferring
glycosyl groups
(GO:0016757)

AT2G37730, AT4G02500, AT5G15740, AT2G29740, AT1G11720, AT5G05870,
AT2G31750, AT2G04560, AT3G46720, AT2G03280, AT3G46680, AT4G26250,
AT4G00300, AT1G60450, AT1G74420, AT4G03340, AT1G09350, AT5G20410,
AT1G54940, AT1G67850, AT5G64740, AT3G21780, AT4G15290, AT1G52420,
AT2G15350, AT4G19900, AT4G13410, AT1G22370, AT1G32930, AT3G27470,
AT3G59100, AT1G71220, AT5G13500, AT3G61130, AT5G41460, AT1G30530,
AT4G04970, AT2G32540, AT3G46970, AT5G03490, AT2G25540, AT1G13250,
AT1G75420, AT2G44660, AT3G46650, AT2G31960, AT5G14860, AT1G27600,
AT2G22930, AT1G70090, AT4G18780, AT3G26370, AT1G38131, AT1G24570,
AT4G15320, AT2G30140, AT2G28080, AT4G17770, AT3G02250, AT1G24170,
AT5G37200, AT1G24070, AT3G11540, AT4G16600, AT2G03220, AT4G15550,
AT2G24630, AT1G20550, AT4G03550, AT3G62660, AT2G32430, AT1G08660,
AT3G29320, AT1G68380, AT1G73810, AT5G01100, AT1G14070, AT5G16190,
AT4G21060, AT2G26100, AT2G19160, AT5G38010, AT3G16520, AT5G59530,
AT1G68020, AT5G49690, AT1G77810, AT1G50580, AT2G21770, AT3G07020,
AT2G23260, AT4G36890, AT5G04480, AT1G78280, AT3G53160, AT1G70290,
AT5G53340, AT2G35650, AT5G12260, AT2G35710, AT2G37980, AT4G09500,
AT2G32610, AT4G14100, AT1G04920, AT5G17040, AT5G47780, AT3G02350,
AT3G21310, AT3G01040, AT1G01570, AT1G11730, AT5G15050, AT3G50760,
AT5G01220, AT4G19460, AT1G70630, AT5G17030, AT1G28240, AT1G38065,
AT1G24100, AT4G12700, AT1G34550, AT1G05570, AT3G21770, AT4G15500,
AT1G74800, AT1G29200, AT3G57200, AT3G01720, AT1G80290, AT1G16900,
AT1G23480, AT1G05150, AT4G11350, AT5G65470, AT1G22340, AT5G11730,
AT1G60140, AT2G20810, AT1G77130, AT4G32110, AT2G36750, AT2G26480,
AT5G03760, AT5G09870, AT3G52060, AT3G03810, AT1G10880, AT1G08040,
AT3G21190, AT1G22460, AT4G39350, AT2G16890, AT4G16590, AT2G43820,
AT3G54100, AT1G67880, AT5G49190, AT1G35510, AT1G11940, AT5G15470,
AT3G43190, AT4G24000, AT2G30150, AT1G26810, AT2G02910, AT1G74380,
AT3G42180, AT2G31790, AT3G57420, AT1G04910, AT1G62330, AT2G01480,
AT2G36780, AT2G03210, AT4G15260, AT5G57270, AT4G26940, AT5G05880,
AT1G71070, AT1G68390, AT1G08990, AT3G11420, AT4G02280, AT2G44500,
AT3G11670, AT3G01620, AT5G46220, AT3G03690, AT4G33330, AT4G32410,
AT5G13000, AT1G18690, AT3G02100, AT3G06440, AT4G31350, AT1G61050,
AT1G05530, AT2G15390, AT2G23210, AT2G30575, AT1G71990, AT4G38310,
AT3G56000, AT5G54010, AT5G25330, AT4G12840, AT5G57500, AT3G25140,
AT5G38460, AT5G03770, AT4G36770, AT5G05170, AT2G41451, AT5G59070,
AT4G27550, AT1G73740, AT5G60700, AT1G78580, AT2G41150, AT1G52630,
AT3G28180, AT1G14020, AT2G29730, AT5G01250, AT4G02130, AT1G11990,
AT3G46700, AT5G22740, AT1G16980, AT1G22400, AT3G14570, AT1G60470,
AT1G63450, AT4G25870, AT3G14960, AT1G05560, AT1G02720, AT2G37580,
AT2G15480, AT1G51630, AT3G45100, AT5G64600, AT3G21760, AT5G63390,
AT1G73160, AT1G20570, AT1G53290, AT5G42660, AT1G07240, AT4G32290,
AT3G29630, AT5G59520, AT3G07170, AT3G30300, AT4G38190, AT1G05280,
AT1G43620, AT1G23870, AT3G08550, AT1G06410, AT4G37690, AT1G03520,
AT1G06780, AT1G32180, AT1G01420, AT5G18480, AT2G36760, AT4G38270,
AT3G55700, AT5G25970, AT3G46670, AT3G24040, AT5G22130, AT4G09630,
AT2G43840, AT2G15370, AT3G15350, AT2G22590, AT2G28310, AT5G54060,
AT1G64910, AT2G18560, AT1G11170, AT2G38650, AT2G38150, AT3G15940,
AT4G27560, AT5G23790, AT5G35570, AT5G44030, AT4G24530, AT4G24010,
AT2G29750, AT2G25300, AT1G73880, AT5G24300, AT3G58790, AT1G06000,
AT4G10120, AT1G76270, AT4G01070, AT1G62305, AT1G55850, AT1G78800,
AT3G19280, AT4G15270, AT3G62720, AT2G36790, AT1G16570, AT2G33100,
AT5G05860, AT1G01390, AT2G32450, AT2G13290, AT3G46690, AT2G36850,

261

Supplemental Table 4.6 (cont’d)

translation
(GO:0006412)

AT3G21800, AT5G05890, AT2G36970, AT4G16710, AT1G14970, AT1G07260,
AT4G16650, AT1G22380, AT4G15280, AT3G03050, AT1G14080, AT4G00550,
AT4G23490, AT4G31590, AT4G23990, AT1G51770, AT4G38240, AT5G22070,
AT1G10280, AT4G17430, AT1G22360, AT2G35100, AT1G33430, AT1G07850,
AT3G06260, AT5G25990, AT1G53040, AT5G67230, AT4G38390, AT1G56600,
AT2G37090, AT3G55830, AT4G27480, AT4G18240, AT1G61240, AT4G18530,
AT2G22900, AT2G39630, AT5G14850, AT1G64920, AT2G20370, AT5G04500,
AT3G50740, AT3G09020, AT3G01180, AT2G46480, AT5G05900, AT1G33250,
AT2G04280, AT5G53990, AT5G07720, AT5G25265, AT4G30060, AT3G56750,
AT4G15480, AT2G15490, AT2G23250, AT3G48820, AT4G31780, AT5G14550,
AT3G55710, AT4G15240, AT1G05680, AT3G21750, AT4G07960, AT1G07250,
AT5G14480, AT1G19300, AT2G36800, AT1G05170, AT1G19710, AT1G05670,
AT4G01210, AT1G32900, AT5G66690, AT2G41770, AT5G65550, AT1G10400,
AT3G27540, AT5G54690, AT3G22250, AT2G36770, AT1G06490, AT4G38500,
AT5G26310, AT2G32530, AT5G16170, AT3G10630, AT3G46660, AT1G12990,
AT2G40190, AT3G07900, AT4G15490, AT2G25260, AT1G51210, AT3G07330,
AT2G18570, AT3G18660, AT1G49710, AT3G04240, AT1G13000, AT2G32620,
AT4G27570, AT5G17050, AT5G16910, AT1G27120, AT5G20280, AT4G08810,
AT2G18700, AT3G26440, AT2G29710, AT5G12890
AT1G14400, AT3G07550, AT3G21860, AT2G03170, AT3G09770, AT2G16920,
AT2G33770, AT2G36370, AT3G06140, AT5G02750, AT2G26000, AT1G80570,
AT4G03510, AT2G01150, AT3G54780, AT4G27470, AT3G29270, AT1G10230,
AT5G06460, AT4G33160, AT1G63900, AT5G43190, AT3G08690, AT2G02760,
AT1G12820, AT2G32950, AT4G34210, AT4G04690, AT1G02860, AT5G49980,
AT1G51550, AT5G59300, AT4G12570, AT5G46210, AT4G10160, AT3G60220,
AT1G36340, AT3G53060, AT2G30110, AT3G26810, AT1G49210, AT4G19700,
AT2G44950, AT5G14420, AT3G13550, AT5G27420, AT3G52560, AT2G04660,
AT4G37890, AT2G30580, AT1G75950, AT4G34470, AT5G05280, AT5G10380,
AT5G65683, AT1G50490, AT1G55860, AT3G60020, AT1G78870, AT5G42190,
AT5G05560, AT2G44330, AT4G03190, AT3G61590, AT4G25230, AT2G22010,
AT5G60710, AT3G21840, AT3G46620, AT3G07370, AT3G19140, AT2G42620,
AT4G28370, AT2G16740, AT5G50430, AT5G02310, AT4G11360, AT1G79810,
AT3G55530, AT3G05870, AT1G77000, AT4G05470, AT5G42200, AT1G57820,
AT5G27920, AT5G51450, AT3G60010, AT5G41700, AT3G17205, AT3G12630,
AT5G38070, AT3G25650, AT2G28830, AT1G75440, AT4G14220, AT5G53300,
AT5G50870, AT3G05545, AT5G02920, AT3G07360, AT1G20140, AT4G27960,
AT2G16810, AT4G05460, AT3G08700, AT3G24515, AT2G39810, AT2G45950,
AT2G35000, AT1G64230, AT3G54850, AT2G38970, AT2G04920, AT2G32790,
AT1G66050, AT1G08050, AT2G25700, AT1G53020, AT3G53410, AT3G04460,
AT1G26830, AT5G67250, AT4G05490, AT1G65430, AT1G30950, AT3G42830,
AT1G69670, AT4G07400, AT3G53090, AT1G70320, AT1G23260, AT3G61415,
AT3G55380, AT4G08980, AT3G17000, AT1G70660, AT5G56150, AT1G29340,
AT5G05080, AT5G25760, AT5G45100, AT1G79380, AT1G45050, AT4G36410,
AT5G57740, AT1G20780, AT2G39940, AT4G23450, AT2G25490, AT1G27910,
AT4G30640, AT5G53840, AT1G22500, AT1G17280, AT3G24800, AT4G28270,
AT1G06770, AT5G49665, AT3G21850, AT5G02880, AT5G18650, AT5G03200,
AT1G10560, AT1G63800, AT5G57360, AT5G42990, SFO1, AT2G18600, AT2G46030,
AT4G33210, AT5G13530, AT3G20060, AT2G35930, AT2G26350, AT3G47990,
AT5G39550, AT2G21950, AT1G15100, AT2G22680, AT1G76920, AT3G18710,
AT4G28890, AT3G01650, AT5G07270, AT3G50080, AT2G18915, AT5G20910,
AT2G42360, AT5G19080, AT5G41340, AT2G36060, AT2G44900, AT4G21070,
AT3G60350, AT2G20160, AT3G46460, AT1G14260, AT3G12775, AT3G05200,
AT3G56580, AT2G42160, AT1G51290, AT1G16890, AT3G21830, AT4G24390,

262

Supplemental Table 4.6 (cont’d)

transporter
activity
(GO:0005215)

AT5G62540, AT2G47700, AT2G03160, AT3G54650, AT5G65200, AT3G21410,
AT3G63530, AT3G06330, AT5G59550, AT5G63970, AT5G53910, AT3G23280,
AT1G21410, AT2G03190, AT3G52450, AT1G12760, AT1G51320, AT1G47056,
AT1G68050, AT3G46510, AT4G15475, AT2G20650, AT3G62980, AT2G03560,
AT5G01720, AT4G08590, AT4G38600, AT2G17020, AT4G02440, AT4G34100
AT2G37730, AT4G02500, AT5G15740, AT2G29740, AT1G11720, AT5G05870,
AT2G31750, AT2G04560, AT3G46720, AT2G03280, AT3G46680, AT4G26250,
AT4G00300, AT1G60450, AT1G74420, AT4G03340, AT1G09350, AT5G20410,
AT1G54940, AT1G67850, AT5G64740, AT3G21780, AT4G15290, AT1G52420,
AT2G15350, AT4G19900, AT4G13410, AT1G22370, AT1G32930, AT3G27470,
AT3G59100, AT1G71220, AT5G13500, AT3G61130, AT5G41460, AT1G30530,
AT4G04970, AT2G32540, AT3G46970, AT5G03490, AT2G25540, AT1G13250,
AT1G75420, AT2G44660, AT3G46650, AT2G31960, AT5G14860, AT1G27600,
AT2G22930, AT1G70090, AT4G18780, AT3G26370, AT1G38131, AT1G24570,
AT4G15320, AT2G30140, AT2G28080, AT4G17770, AT3G02250, AT1G24170,
AT5G37200, AT1G24070, AT3G11540, AT4G16600, AT2G03220, AT4G15550,
AT2G24630, AT1G20550, AT4G03550, AT3G62660, AT2G32430, AT1G08660,
AT3G29320, AT1G68380, AT1G73810, AT5G01100, AT1G14070, AT5G16190,
AT4G21060, AT2G26100, AT2G19160, AT5G38010, AT3G16520, AT5G59530,
AT1G68020, AT5G49690, AT1G77810, AT1G50580, AT2G21770, AT3G07020,
AT2G23260, AT4G36890, AT5G04480, AT1G78280, AT3G53160, AT1G70290,
AT5G53340, AT2G35650, AT5G12260, AT2G35710, AT2G37980, AT4G09500,
AT2G32610, AT4G14100, AT1G04920, AT5G17040, AT5G47780, AT3G02350,
AT3G21310, AT3G01040, AT1G01570, AT1G11730, AT5G15050, AT3G50760,
AT5G01220, AT4G19460, AT1G70630, AT5G17030, AT1G28240, AT1G38065,
AT1G24100, AT4G12700, AT1G34550, AT1G05570, AT3G21770, AT4G15500,
AT1G74800, AT1G29200, AT3G57200, AT3G01720, AT1G80290, AT1G16900,
AT1G23480, AT1G05150, AT4G11350, AT5G65470, AT1G22340, AT5G11730,
AT1G60140, AT2G20810, AT1G77130, AT4G32110, AT2G36750, AT2G26480,
AT5G03760, AT5G09870, AT3G52060, AT3G03810, AT1G10880, AT1G08040,
AT3G21190, AT1G22460, AT4G39350, AT2G16890, AT4G16590, AT2G43820,
AT3G54100, AT1G67880, AT5G49190, AT1G35510, AT1G11940, AT5G15470,
AT3G43190, AT4G24000, AT2G30150, AT1G26810, AT2G02910, AT1G74380,
AT3G42180, AT2G31790, AT3G57420, AT1G04910, AT1G62330, AT2G01480,
AT2G36780, AT2G03210, AT4G15260, AT5G57270, AT4G26940, AT5G05880,
AT1G71070, AT1G68390, AT1G08990, AT3G11420, AT4G02280, AT2G44500,
AT3G11670, AT3G01620, AT5G46220, AT3G03690, AT4G33330, AT4G32410,
AT5G13000, AT1G18690, AT3G02100, AT3G06440, AT4G31350, AT1G61050,
AT1G05530, AT2G15390, AT2G23210, AT2G30575, AT1G71990, AT4G38310,
AT3G56000, AT5G54010, AT5G25330, AT4G12840, AT5G57500, AT3G25140,
AT5G38460, AT5G03770, AT4G36770, AT5G05170, AT2G41451, AT5G59070,
AT4G27550, AT1G73740, AT5G60700, AT1G78580, AT2G41150, AT1G52630,
AT3G28180, AT1G14020, AT2G29730, AT5G01250, AT4G02130, AT1G11990,
AT3G46700, AT5G22740, AT1G16980, AT1G22400, AT3G14570, AT1G60470,
AT1G63450, AT4G25870, AT3G14960, AT1G05560, AT1G02720, AT2G37580,
AT2G15480, AT1G51630, AT3G45100, AT5G64600, AT3G21760, AT5G63390,
AT1G73160, AT1G20570, AT1G53290, AT5G42660, AT1G07240, AT4G32290,
AT3G29630, AT5G59520, AT3G07170, AT3G30300, AT4G38190, AT1G05280,
AT1G43620, AT1G23870, AT3G08550, AT1G06410, AT4G37690, AT1G03520,
AT1G06780, AT1G32180, AT1G01420, AT5G18480, AT2G36760, AT4G38270,
AT3G55700, AT5G25970, AT3G46670, AT3G24040, AT5G22130, AT4G09630,
AT2G43840, AT2G15370, AT3G15350, AT2G22590, AT2G28310, AT5G54060,
AT1G64910, AT2G18560, AT1G11170, AT2G38650, AT2G38150, AT3G15940,

263

Supplemental Table 4.6 (cont’d)

ubiquitin-protein
transferase
activity
(GO:0004842)

AT4G27560, AT5G23790, AT5G35570, AT5G44030, AT4G24530, AT4G24010,
AT2G29750, AT2G25300, AT1G73880, AT5G24300, AT3G58790, AT1G06000,
AT4G10120, AT1G76270, AT4G01070, AT1G62305, AT1G55850, AT1G78800,
AT3G19280, AT4G15270, AT3G62720, AT2G36790, AT1G16570, AT2G33100,
AT5G05860, AT1G01390, AT2G32450, AT2G13290, AT3G46690, AT2G36850,
AT3G21800, AT5G05890, AT2G36970, AT4G16710, AT1G14970, AT1G07260,
AT4G16650, AT1G22380, AT4G15280, AT3G03050, AT1G14080, AT4G00550,
AT4G23490, AT4G31590, AT4G23990, AT1G51770, AT4G38240, AT5G22070,
AT1G10280, AT4G17430, AT1G22360, AT2G35100, AT1G33430, AT1G07850,
AT3G06260, AT5G25990, AT1G53040, AT5G67230, AT4G38390, AT1G56600,
AT2G37090, AT3G55830, AT4G27480, AT4G18240, AT1G61240, AT4G18530,
AT2G22900, AT2G39630, AT5G14850, AT1G64920, AT2G20370, AT5G04500,
AT3G50740, AT3G09020, AT3G01180, AT2G46480, AT5G05900, AT1G33250,
AT2G04280, AT5G53990, AT5G07720, AT5G25265, AT4G30060, AT3G56750,
AT4G15480, AT2G15490, AT2G23250, AT3G48820, AT4G31780, AT5G14550,
AT3G55710, AT4G15240, AT1G05680, AT3G21750, AT4G07960, AT1G07250,
AT5G14480, AT1G19300, AT2G36800, AT1G05170, AT1G19710, AT1G05670,
AT4G01210, AT1G32900, AT5G66690, AT2G41770, AT5G65550, AT1G10400,
AT3G27540, AT5G54690, AT3G22250, AT2G36770, AT1G06490, AT4G38500,
AT5G26310, AT2G32530, AT5G16170, AT3G10630, AT3G46660, AT1G12990,
AT2G40190, AT3G07900, AT4G15490, AT2G25260, AT1G51210, AT3G07330,
AT2G18570, AT3G18660, AT1G49710, AT3G04240, AT1G13000, AT2G32620,
AT4G27570, AT5G17050, AT5G16910, AT1G27120, AT5G20280, AT4G08810,
AT2G18700, AT3G26440, AT2G29710, AT5G12890
AT1G14400, AT3G07550, AT3G21860, AT2G03170, AT3G09770, AT2G16920,
AT2G33770, AT2G36370, AT3G06140, AT5G02750, AT2G26000, AT1G80570,
AT4G03510, AT2G01150, AT3G54780, AT4G27470, AT3G29270, AT1G10230,
AT5G06460, AT4G33160, AT1G63900, AT5G43190, AT3G08690, AT2G02760,
AT1G12820, AT2G32950, AT4G34210, AT4G04690, AT1G02860, AT5G49980,
AT1G51550, AT5G59300, AT4G12570, AT5G46210, AT4G10160, AT3G60220,
AT1G36340, AT3G53060, AT2G30110, AT3G26810, AT1G49210, AT4G19700,
AT2G44950, AT5G14420, AT3G13550, AT5G27420, AT3G52560, AT2G04660,
AT4G37890, AT2G30580, AT1G75950, AT4G34470, AT5G05280, AT5G10380,
AT5G65683, AT1G50490, AT1G55860, AT3G60020, AT1G78870, AT5G42190,
AT5G05560, AT2G44330, AT4G03190, AT3G61590, AT4G25230, AT2G22010,
AT5G60710, AT3G21840, AT3G46620, AT3G07370, AT3G19140, AT2G42620,
AT4G28370, AT2G16740, AT5G50430, AT5G02310, AT4G11360, AT1G79810,
AT3G55530, AT3G05870, AT1G77000, AT4G05470, AT5G42200, AT1G57820,
AT5G27920, AT5G51450, AT3G60010, AT5G41700, AT3G17205, AT3G12630,
AT5G38070, AT3G25650, AT2G28830, AT1G75440, AT4G14220, AT5G53300,
AT5G50870, AT3G05545, AT5G02920, AT3G07360, AT1G20140, AT4G27960,
AT2G16810, AT4G05460, AT3G08700, AT3G24515, AT2G39810, AT2G45950,
AT2G35000, AT1G64230, AT3G54850, AT2G38970, AT2G04920, AT2G32790,
AT1G66050, AT1G08050, AT2G25700, AT1G53020, AT3G53410, AT3G04460,
AT1G26830, AT5G67250, AT4G05490, AT1G65430, AT1G30950, AT3G42830,
AT1G69670, AT4G07400, AT3G53090, AT1G70320, AT1G23260, AT3G61415,
AT3G55380, AT4G08980, AT3G17000, AT1G70660, AT5G56150, AT1G29340,
AT5G05080, AT5G25760, AT5G45100, AT1G79380, AT1G45050, AT4G36410,
AT5G57740, AT1G20780, AT2G39940, AT4G23450, AT2G25490, AT1G27910,
AT4G30640, AT5G53840, AT1G22500, AT1G17280, AT3G24800, AT4G28270,
AT1G06770, AT5G49665, AT3G21850, AT5G02880, AT5G18650, AT5G03200,
AT1G10560, AT1G63800, AT5G57360, AT5G42990, SFO1, AT2G18600, AT2G46030,
AT4G33210, AT5G13530, AT3G20060, AT2G35930, AT2G26350, AT3G47990,

264

Supplemental Table 4.6 (cont’d)

zinc ion binding
(GO:0008270)

AT5G39550, AT2G21950, AT1G15100, AT2G22680, AT1G76920, AT3G18710,
AT4G28890, AT3G01650, AT5G07270, AT3G50080, AT2G18915, AT5G20910,
AT2G42360, AT5G19080, AT5G41340, AT2G36060, AT2G44900, AT4G21070,
AT3G60350, AT2G20160, AT3G46460, AT1G14260, AT3G12775, AT3G05200,
AT3G56580, AT2G42160, AT1G51290, AT1G16890, AT3G21830, AT4G24390,
AT5G62540, AT2G47700, AT2G03160, AT3G54650, AT5G65200, AT3G21410,
AT3G63530, AT3G06330, AT5G59550, AT5G63970, AT5G53910, AT3G23280,
AT1G21410, AT2G03190, AT3G52450, AT1G12760, AT1G51320, AT1G47056,
AT1G68050, AT3G46510, AT4G15475, AT2G20650, AT3G62980, AT2G03560,
AT5G01720, AT4G08590, AT4G38600, AT2G17020, AT4G02440, AT4G34100
AT1G14400, AT3G07550, AT3G21860, AT2G03170, AT3G09770, AT2G16920,
AT2G33770, AT2G36370, AT3G06140, AT5G02750, AT2G26000, AT1G80570,
AT4G03510, AT2G01150, AT3G54780, AT4G27470, AT3G29270, AT1G10230,
AT5G06460, AT4G33160, AT1G63900, AT5G43190, AT3G08690, AT2G02760,
AT1G12820, AT2G32950, AT4G34210, AT4G04690, AT1G02860, AT5G49980,
AT1G51550, AT5G59300, AT4G12570, AT5G46210, AT4G10160, AT3G60220,
AT1G36340, AT3G53060, AT2G30110, AT3G26810, AT1G49210, AT4G19700,
AT2G44950, AT5G14420, AT3G13550, AT5G27420, AT3G52560, AT2G04660,
AT4G37890, AT2G30580, AT1G75950, AT4G34470, AT5G05280, AT5G10380,
AT5G65683, AT1G50490, AT1G55860, AT3G60020, AT1G78870, AT5G42190,
AT5G05560, AT2G44330, AT4G03190, AT3G61590, AT4G25230, AT2G22010,
AT5G60710, AT3G21840, AT3G46620, AT3G07370, AT3G19140, AT2G42620,
AT4G28370, AT2G16740, AT5G50430, AT5G02310, AT4G11360, AT1G79810,
AT3G55530, AT3G05870, AT1G77000, AT4G05470, AT5G42200, AT1G57820,
AT5G27920, AT5G51450, AT3G60010, AT5G41700, AT3G17205, AT3G12630,
AT5G38070, AT3G25650, AT2G28830, AT1G75440, AT4G14220, AT5G53300,
AT5G50870, AT3G05545, AT5G02920, AT3G07360, AT1G20140, AT4G27960,
AT2G16810, AT4G05460, AT3G08700, AT3G24515, AT2G39810, AT2G45950,
AT2G35000, AT1G64230, AT3G54850, AT2G38970, AT2G04920, AT2G32790,
AT1G66050, AT1G08050, AT2G25700, AT1G53020, AT3G53410, AT3G04460,
AT1G26830, AT5G67250, AT4G05490, AT1G65430, AT1G30950, AT3G42830,
AT1G69670, AT4G07400, AT3G53090, AT1G70320, AT1G23260, AT3G61415,
AT3G55380, AT4G08980, AT3G17000, AT1G70660, AT5G56150, AT1G29340,
AT5G05080, AT5G25760, AT5G45100, AT1G79380, AT1G45050, AT4G36410,
AT5G57740, AT1G20780, AT2G39940, AT4G23450, AT2G25490, AT1G27910,
AT4G30640, AT5G53840, AT1G22500, AT1G17280, AT3G24800, AT4G28270,
AT1G06770, AT5G49665, AT3G21850, AT5G02880, AT5G18650, AT5G03200,
AT1G10560, AT1G63800, AT5G57360, AT5G42990, SFO1, AT2G18600, AT2G46030,
AT4G33210, AT5G13530, AT3G20060, AT2G35930, AT2G26350, AT3G47990,
AT5G39550, AT2G21950, AT1G15100, AT2G22680, AT1G76920, AT3G18710,
AT4G28890, AT3G01650, AT5G07270, AT3G50080, AT2G18915, AT5G20910,
AT2G42360, AT5G19080, AT5G41340, AT2G36060, AT2G44900, AT4G21070,
AT3G60350, AT2G20160, AT3G46460, AT1G14260, AT3G12775, AT3G05200,
AT3G56580, AT2G42160, AT1G51290, AT1G16890, AT3G21830, AT4G24390,
AT5G62540, AT2G47700, AT2G03160, AT3G54650, AT5G65200, AT3G21410,
AT3G63530, AT3G06330, AT5G59550, AT5G63970, AT5G53910, AT3G23280,
AT1G21410, AT2G03190, AT3G52450, AT1G12760, AT1G51320, AT1G47056,
AT1G68050, AT3G46510, AT4G15475, AT2G20650, AT3G62980, AT2G03560,
AT5G01720, AT4G08590, AT4G38600, AT2G17020, AT4G02440, AT4G34100

265

Supplemental Table 4.7 TF genes belonging to each TF family in A. thaliana
TF Family

TFs

AP2

AT1G64380, AT5G47220, AT3G20310, AT4G18450, AT2G35700, AT4G28140, AT5G57390,
AT1G25560, AT1G22810, AT2G20350, AT1G12610, AT1G71520, AT4G25480, AT2G23340,
AT1G46768, AT3G25730, AT5G65510, AT2G40340, AT5G65130, AT5G18450, AT5G21960,
AT5G61600, AT1G36060, AT4G25470, AT1G21910, AT4G36900, AT1G79700, AT4G39780,
AT3G50260, AT1G80580, AT1G28160, AT4G11140, AT1G12890, AT5G53290, AT3G61630,
AT5G64750, AT5G25190, AT3G60490, AT5G51990, AT1G12630, AT2G33710, AT2G44940,
AT5G67000, AT2G39250, AT1G71130, AT1G06160, AT3G16770, AT1G51120, AT5G17430,
AT1G72570, AT1G04370, AT2G31230, AT1G19210, AT3G54990, AT2G36450, AT5G61890,
AT5G44210, AT1G50640, AT4G34410, AT5G43410, AT1G77640, AT4G36920, AT2G25820,
AT1G78080, AT1G43160, AT5G52020, AT1G22985, AT2G20880, AT3G20840, AT2G46310,
AT4G17500, AT1G63030, AT4G13040, AT5G18560, AT4G16750, AT2G44840, AT4G25490,
AT5G10510, AT1G44830, AT5G13330, AT1G50680, AT4G32800, AT1G53910, AT3G23240,
AT4G23750, AT2G41710, AT4G13620, AT5G13910, AT3G23220, AT5G25810, AT2G47520,
AT2G22200, AT5G51190, AT3G54320, AT1G51190, AT1G24590, AT3G11020, AT1G28360,
AT3G15210, AT4G37750, AT5G07580, AT5G11190, AT1G25470, AT1G22190, AT1G72360,
AT1G75490, AT1G49120, AT5G60120, AT3G23230, AT4G27950, AT3G57600, AT5G47230,
AT1G15360, AT3G25890, AT1G28370, AT1G53170, AT5G67190, AT2G28550, AT5G25390,
AT5G19790, AT1G77200, AT1G74930, AT1G68550, AT1G33760, AT5G11590, AT3G16280,
AT1G03800, AT1G68840, AT1G12980, AT5G07310, AT4G31060, AT2G40220, AT1G01250,
AT5G61590, AT1G16060, AT4G06746, AT5G50080, AT5G67180, AT4G17490, AT3G14230,
AT5G67010, AT2G38340, AT5G05410, AT2G40350, AT1G13260, AT1G71450
AT1G35540, AT2G30470, AT2G36080, AT5G32460, AT2G24680, AT4G31690, AT1G43950,
AT1G19220, AT4G31660, AT4G03170, AT3G25730, AT4G31610, AT5G62000, AT3G11580,
AT2G28350, AT3G06220, AT1G34310, AT5G60142, AT1G35240, AT5G60140, AT3G06160,
AT4G31640, AT1G30330, AT1G34390, AT1G34410, AT2G46870, AT5G57720, AT3G61830,
AT1G20600, AT5G06250, AT1G51120, AT1G19850, AT4G31650, AT2G24650, AT3G17010,
AT4G01580, AT3G53310, AT5G20730, AT5G18000, AT5G42700, AT2G33860, AT1G59750,
AT4G31620, AT1G25560, AT5G25470, AT5G58280, AT5G25475, AT2G24700, AT1G01030,
AT2G35310, AT5G09780, AT4G34400, AT1G50680, AT2G24645, AT2G16210, AT4G00260,
AT4G21550, AT4G01500, AT1G49480, AT4G33280, AT1G34170, AT1G16640, AT4G31630,
AT1G77850, AT3G18990, AT4G31615, AT2G33720, AT4G30080, AT3G46770, AT4G23980,
AT3G26790, AT3G18960, AT1G08985, AT5G18090, AT1G49475, AT1G26680, AT3G19184,
AT2G24681, AT3G61970, AT5G60130, AT1G05930, AT3G24650, AT1G68840, AT5G66980,
AT5G37020, AT4G32010, AT2G24690, AT4G31680, AT2G24696, AT1G28300, AT5G60450,
AT1G35520, AT1G13260, AT2G46530
AT5G46760, AT4G09820, AT1G32640, AT4G00870, AT2G31280, AT2G16910, AT4G16430,
AT4G17880, AT1G10610, AT1G63650, AT2G46510, AT5G41315, AT1G01260, AT4G00480,
AT5G46830
AT3G54620, AT3G19290, AT5G11260, AT2G17770, AT1G77920, AT5G24800, AT3G51960,
AT3G44460, AT2G16770, AT1G32150, AT4G35900, AT3G30530, AT2G18160, AT1G06850,
AT1G75390, AT2G46270, AT2G21230, AT5G06839, AT1G22070, AT5G06950, AT4G34590,
AT2G40620, AT5G60830, AT4G37730, AT3G49760, AT4G36730, AT4G35040, AT2G35530,
AT2G41070, AT3G62420, AT2G42380, AT4G34000, AT4G38900, AT5G28770, AT3G56660,
AT2G04038, AT4G02640, AT5G42910, AT1G08320, AT3G17609, AT5G07160, AT2G31370,
AT5G38800, AT5G08141, AT5G49450, AT1G68880, AT5G15830, AT3G58120, AT2G12940,
AT5G44080, AT2G13150, AT3G10800, AT1G42990, AT2G40950, AT5G06960, AT3G56850,
AT1G68640, AT1G13600, AT2G22850, AT4G01120, AT1G43700, AT5G10030, AT1G49720,
AT5G65210, AT1G06070, AT3G12250, AT2G36270, AT1G03970, AT1G59530

B3

bHLHMYC N
bZIP 1

266

Supplemental Table 4.7 (cont’d)
bZIP 2

bZIP C
CBFB
NFYA
E2F TDP
EIN3
FAR1
GAGA
GATA

HALZ
HLH

Homeobox
KN

AT3G54620, AT3G19290, AT5G11260, AT2G17770, AT5G24800, AT3G51960, AT3G44460,
AT2G16770, AT1G32150, AT4G35900, AT3G30530, AT2G18160, AT1G06850, AT1G75390,
AT2G46270, AT2G21230, AT5G06839, AT4G38900, AT5G06950, AT2G22850, AT2G40620,
AT5G60830, AT4G37730, AT3G49760, AT5G44080, AT4G36730, AT4G35040, AT2G35530,
AT2G41070, AT3G62420, AT2G42380, AT4G34000, AT5G07160, AT5G28770, AT3G56660,
AT2G04038, AT4G02640, AT5G42910, AT1G08320, AT3G17609, AT2G31370, AT5G38800,
AT5G08141, AT5G49450, AT2G12900, AT1G68880, AT5G15830, AT3G58120, AT2G12940,
AT4G34590, AT2G13150, AT3G10800, AT1G42990, AT2G40950, AT5G06960, AT3G56850,
AT1G68640, AT1G13600, AT1G45249, AT4G01120, AT1G43700, AT5G10030, AT1G49720,
AT5G65210, AT1G06070, AT3G12250, AT2G36270, AT1G03970, AT1G59530
AT4G02640, AT5G24800, AT5G28770, AT3G54620
AT3G05690, AT1G17590, AT3G20910, AT1G30500, AT1G72830, AT2G34720, AT3G14020,
AT1G54160, AT5G06510, AT5G12840
AT3G48160, AT5G14960, AT2G36010, AT1G47870, AT5G02470, AT3G01330, AT5G22220,
AT5G03415
AT5G21120, AT3G20770, AT2G27050, AT5G65100, AT1G73730, AT5G10120
AT5G18960, AT1G52520, AT5G28530, AT2G32250, AT4G19990, AT3G07500, AT1G10240,
AT1G76320, AT3G59470, AT4G12850, AT4G15090, AT2G43280, AT2G27110, AT4G38180,
AT3G06250, AT3G22170, AT1G80010
AT2G01930, AT2G35550, AT1G68120, AT1G14685, AT4G38910, AT5G42520, AT2G21240
AT1G51600, AT5G56860, AT3G24050, AT3G60530, AT3G21175, AT1G08010, AT5G47140,
AT4G34680, AT3G45170, AT5G66320, AT2G28340, AT3G50870, AT5G26930, AT4G36240,
AT3G16870, AT5G25830, AT3G06740, AT1G08000, AT3G54810, AT4G24470, AT4G17570,
AT3G20750, AT2G45050, AT3G51080, AT2G18380, AT4G36620, AT4G16141, AT4G26150,
AT4G32890, AT5G49300
AT5G65310, AT2G22430, AT3G60390, AT2G01430, AT5G06710, AT2G46680, AT3G61890,
AT3G01220, AT2G22800, AT3G01470, AT1G27045, AT4G37790, AT4G17460, AT5G47370,
AT4G40060, AT1G70920, AT1G26960, AT5G15150, AT1G69780, AT2G44910, AT4G16780
AT5G61270, AT5G37800, AT1G01260, AT4G25400, AT5G43650, AT2G43010, AT2G42300,
AT3G26744, AT1G05805, AT1G51140, AT4G02590, AT1G59640, AT5G51780, AT1G74500,
AT1G66470, AT1G71200, AT5G65320, AT3G56970, AT5G58010, AT4G21330, AT4G16430,
AT1G72210, AT1G10610, AT2G31210, AT5G41315, AT2G31215, AT3G56980, AT3G23690,
AT4G33880, AT1G10120, AT5G46690, AT5G43175, AT2G46510, AT3G19860, AT5G01310,
AT3G23210, AT2G42280, AT3G47710, AT5G08130, AT4G00050, AT4G28790, AT2G31220,
AT5G62610, AT3G59060, AT2G41130, AT2G46970, AT1G18400, AT3G25710, AT1G68920,
AT4G09820, AT5G51790, AT5G48560, AT3G24140, AT2G22770, AT2G24260, AT1G22490,
AT1G06170, AT2G14760, AT1G68810, AT1G26260, AT1G35460, AT5G46760, AT4G36930,
AT2G22760, AT5G54680, AT5G65640, AT1G03040, AT3G61950, AT3G56770, AT2G20180,
AT1G12860, AT4G30980, AT5G67060, AT2G46810, AT1G26945, AT2G41240, AT3G07340,
AT3G50330, AT2G40200, AT1G12540, AT5G57150, AT5G50915, AT4G17880, AT2G28160,
AT1G69010, AT2G22750, AT1G27740, AT4G34530, AT1G02340, AT4G14410, AT4G37850,
AT1G09530, AT3G62090, AT1G32640, AT4G00870, AT5G10570, AT5G04150, AT3G06120,
AT4G00120, AT4G09180, AT3G21330, AT4G36540, AT4G28800, AT5G67110, AT5G56960,
AT1G49770, AT5G38860, AT2G16910, AT2G18300, AT4G20970, AT2G43140, AT5G09750,
AT1G68240, AT5G53210, AT5G46830, AT1G62975, AT4G29930, AT4G25410, AT4G28811,
AT1G25330, AT4G28815, AT1G63650, AT4G38070, AT1G51070, AT4G01460, AT1G73830
AT2G27220, AT1G75430, AT1G62990, AT1G75410, AT2G23760, AT4G36870, AT4G32040,
AT1G62360, AT2G35940, AT5G02030, AT5G11060, AT1G70510, AT4G32980, AT1G23380,
AT4G25530, AT2G16400, AT5G25220, AT2G27990, AT4G34610, AT5G41410, AT1G19700,
AT4G08150

267

Supplemental Table 4.7 (cont’d)
HSF DNAbind
K-box

KNOX1
KNOX2
MADF
DNA bdg
MEKHLA

AT2G27220, AT1G75430, AT1G62990, AT1G75410, AT2G23760, AT4G36870, AT4G32040,
AT1G62360, AT2G35940, AT5G02030, AT5G11060, AT1G70510, AT4G32980, AT1G23380,
AT4G25530, AT2G16400, AT5G25220, AT2G27990, AT4G34610, AT5G41410, AT1G19700,
AT4G08150
AT5G20240, AT1G31140, AT2G22540, AT5G65060, AT4G22950, AT5G51860, AT3G57230,
AT4G24540, AT5G65080, AT3G02310, AT5G60910, AT1G69120, AT2G45650, AT3G57390,
AT1G24260, AT1G77080, AT4G18960, AT5G10140, AT3G61120, AT4G11880, AT5G65050,
AT5G51870, AT1G26310, AT4G37940, AT5G23260, AT2G22630, AT4G09960, AT5G62165,
AT2G14210, AT3G54340, AT5G65070, AT2G45660, AT3G58780, AT5G15800, AT2G03710,
AT2G42830, AT1G71692, AT5G13790, AT3G30260
AT4G32040, AT1G70510, AT5G25220, AT5G11060, AT1G62990, AT1G23380, AT1G62360,
AT4G08150
AT4G32040, AT1G70510, AT5G25220, AT5G11060, AT1G62990, AT1G23380, AT1G62360,
AT4G08150
AT1G33240, AT3G10030, AT3G24490, AT1G76880, AT1G76890, AT2G44730
AT1G30490, AT2G34710, AT5G60690, AT4G32880, AT1G52150

MFMR

AT4G01120, AT2G46270, AT2G35530, AT4G36730, AT1G32150

Myb CC
LHEQLE

AT3G12730, AT3G13040, AT4G13640, AT4G28610, AT5G29000, AT3G04030, AT3G04450,
AT5G06800, AT5G18240, AT5G45580, AT3G24120, AT1G69580, AT2G20400, AT1G79430,
AT2G01060
AT1G31310, AT1G33240, AT3G58630, AT1G76880, AT2G44730, AT2G35640, AT1G21200,
AT2G33550, AT3G10030, AT3G24490, AT3G24860, AT1G76870, AT2G38250, AT5G63420,
AT5G01380, AT5G03680, AT5G05550, AT1G13450, AT5G47660, AT1G54060, AT1G76890,
AT3G11100, AT3G14180, AT3G10040, AT5G28300, AT3G54390, AT4G31270, AT3G10000,
AT3G25990
AT1G33240, AT3G25990, AT1G76890, AT1G13450

Myb DNAbind 4

Myb DNAbind 5
Myb DNAbind 6

AT3G50060, AT3G24310, AT3G09370, AT4G38620, AT5G40360, AT2G37630, AT4G25560,
AT5G65790, AT1G56160, AT3G01530, AT5G12870, AT4G05100, AT4G32730, AT1G79180,
AT3G13540, AT1G48000, AT3G01140, AT4G17785, AT4G37780, AT2G31180, AT3G46130,
AT5G14340, AT1G72740, AT3G06490, AT3G52250, AT3G27810, AT1G76890, AT4G13480,
AT2G25230, AT5G56110, AT4G21440, AT3G28470, AT5G07700, AT5G14750, AT1G66370,
AT4G00540, AT1G18570, AT1G74430, AT4G37260, AT3G49690, AT5G07690, AT5G54230,
AT4G33450, AT3G11450, AT2G36890, AT3G27785, AT2G13960, AT5G59780, AT2G32460,
AT3G55730, AT3G48920, AT1G08810, AT3G13890, AT2G39880, AT3G10113, AT3G12820,
AT1G71030, AT2G47190, AT4G34990, AT1G25340, AT4G22680, AT2G47460, AT1G15720,
AT5G40350, AT4G28110, AT1G18960, AT5G52600, AT5G45420, AT3G11280, AT1G18710,
AT1G74080, AT3G12720, AT5G41020, AT1G26780, AT3G29020, AT5G67300, AT5G10280,
AT5G55020, AT5G61420, AT3G53200, AT5G39700, AT3G02940, AT3G09230, AT1G22640,
AT5G62470, AT5G57620, AT5G11050, AT3G27920, AT5G40430, AT5G35550, AT1G34670,
AT3G61250, AT1G17520, AT3G49850, AT5G17800, AT5G60890, AT5G01200, AT1G74650,
AT1G69560, AT3G11440, AT2G26950, AT5G65230, AT3G28910, AT1G66230, AT1G35515,
AT1G73410, AT5G58340, AT5G23000, AT5G11510, AT3G47600, AT5G40330, AT4G01680,
AT2G16720, AT3G60460, AT5G62320, AT4G09460, AT1G14350, AT5G06100, AT3G23250,
AT1G09540, AT1G57560, AT2G38090, AT1G72650, AT2G46410, AT1G18330, AT5G16600,
AT5G15310, AT5G04760, AT4G12350, AT2G23290, AT5G05790, AT4G18770, AT5G58850,
AT1G16490, AT1G56650, AT1G58220, AT5G02320, AT5G06110, AT3G30210, AT1G17950,
AT1G66380, AT5G16770, AT3G08500, AT5G49330, AT5G49620, AT1G06180, AT1G68320,
AT5G58900, AT4G26930, AT3G18100, AT5G52260, AT5G26660, AT1G63910, AT2G02820,
AT1G09770, AT2G26960, AT5G67580, AT3G62610, AT1G66390

268

Supplemental Table 4.7 (cont’d)
Myb DNAbinding

NAM

AT3G50060, AT2G25180, AT3G24310, AT3G09370, AT4G38620, AT5G40360, AT2G37630,
AT3G57980, AT5G05790, AT2G44430, AT5G65790, AT1G56160, AT3G01530, AT5G12870,
AT4G05100, AT1G01380, AT4G32730, AT1G79180, AT5G02840, AT3G13540, AT3G25790,
AT4G18020, AT3G10590, AT4G09450, AT5G59570, AT3G04450, AT1G48000, AT3G01140,
AT4G17785, AT4G37780, AT1G66230, AT2G31180, AT3G46130, AT5G14340, AT4G13640,
AT1G72740, AT2G18328, AT3G52250, AT1G70000, AT1G06910, AT3G27810, AT4G13480,
AT2G40970, AT4G04580, AT2G25230, AT1G66380, AT5G39700, AT5G56110, AT4G21440,
AT3G28470, AT5G07700, AT5G08520, AT4G36570, AT1G66370, AT4G25560, AT4G00540,
AT1G18570, AT1G22640, AT4G37260, AT3G49690, AT5G07690, AT5G54230, AT4G33450,
AT3G11450, AT2G36890, AT3G27785, AT2G13960, AT5G59780, AT3G08500, AT2G32460,
AT5G26660, AT1G17460, AT5G58080, AT1G69580, AT3G55730, AT5G18240, AT5G53200,
AT3G12730, AT3G48920, AT4G16110, AT1G08810, AT3G13890, AT2G39880, AT3G10113,
AT3G12820, AT1G71030, AT5G06100, AT2G47190, AT4G34990, AT1G25340, AT3G10760,
AT4G22680, AT2G47460, AT4G37180, AT1G15720, AT5G40350, AT1G09710, AT3G53790,
AT5G67300, AT5G41020, AT4G28110, AT1G18960, AT5G52600, AT4G39250, AT3G11280,
AT2G06020, AT1G79430, AT2G01060, AT1G74080, AT3G12720, AT5G44190, AT4G31920,
AT5G16770, AT1G26780, AT3G29020, AT2G42660, AT5G61620, AT3G10580, AT5G10280,
AT5G55020, AT1G18330, AT4G17695, AT3G53200, AT2G38300, AT3G02940, AT3G09230,
AT5G62470, AT5G57620, AT5G11050, AT3G60460, AT5G23650, AT3G46640, AT3G27920,
AT2G30432, AT5G40430, AT5G35550, AT5G45420, AT3G62670, AT1G34670, AT3G61250,
AT1G17520, AT1G74430, AT3G49850, AT5G17800, AT5G60890, AT4G28610, AT5G29000,
AT1G18710, AT5G01200, AT1G74650, AT1G13300, AT1G68670, AT1G69560, AT3G11440,
AT2G26950, AT2G40260, AT4G39160, AT3G28910, AT1G49950, AT5G05090, AT1G35515,
AT1G73410, AT3G13040, AT5G37260, AT5G23000, AT1G49010, AT3G47600, AT5G40330,
AT3G04030, AT4G01680, AT2G16720, AT2G30420, AT5G62320, AT2G30424, AT3G24120,
AT4G09460, AT3G21430, AT4G01280, AT5G07210, AT5G47390, AT1G14350, AT1G14600,
AT3G23250, AT2G20570, AT1G09540, AT1G57560, AT5G11510, AT1G19000, AT2G38090,
AT1G72650, AT2G46410, AT2G03500, AT5G16600, AT5G15310, AT3G09600, AT5G17300,
AT5G04760, AT5G65230, AT5G56840, AT1G01520, AT3G16350, AT4G12350, AT2G01760,
AT2G23290, AT5G14750, AT5G42630, AT1G67710, AT4G18770, AT1G32240, AT5G58340,
AT5G58850, AT1G74840, AT1G16490, AT3G60110, AT1G56650, AT1G58220, AT5G02320,
AT5G06110, AT3G30210, AT2G20400, AT1G17950, AT1G25550, AT5G16560, AT3G16857,
AT1G49560, AT3G06490, AT5G49330, AT5G49620, AT5G06800, AT2G36960, AT1G06180,
AT1G68320, AT1G01060, AT5G58900, AT5G61420, AT4G26930, AT3G18100, AT5G52260,
AT4G01060, AT1G63910, AT2G02820, AT1G09770, AT5G52660, AT2G26960, AT2G46830,
AT5G45580, AT5G67580, AT3G62610, AT2G02060, AT1G66390, AT2G42150
AT3G49530, AT5G64530, AT5G39690, AT5G63790, AT3G61910, AT3G15510, AT5G62380,
AT5G04400, AT1G79580, AT1G60350, AT4G36160, AT5G56620, AT3G01600, AT2G46770,
AT3G55210, AT1G60280, AT5G46590, AT5G13180, AT2G43000, AT2G33480, AT1G03490,
AT5G09330, AT2G27300, AT1G19040, AT4G01550, AT1G54330, AT3G10500, AT5G07680,
AT1G71930, AT3G56530, AT1G02250, AT5G66300, AT4G27410, AT5G50820, AT1G76420,
AT3G04070, AT4G01520, AT1G32510, AT4G17980, AT3G56520, AT3G44350, AT1G60380,
AT5G39820, AT1G25580, AT5G22290, AT3G15170, AT1G60300, AT5G24590, AT3G04060,
AT5G18270, AT5G64060, AT5G14000, AT3G18400, AT5G08790, AT1G64105, AT1G02230,
AT3G29035, AT3G04410, AT5G41090, AT3G44290, AT1G65910, AT4G10350, AT5G39610,
AT3G10480, AT4G29230, AT4G28500, AT1G12260, AT1G56010, AT1G52890, AT2G24430,
AT1G28470, AT1G01010, AT1G69490, AT1G33280, AT5G61430, AT1G02220, AT2G17040,
AT5G14490, AT3G12910, AT1G60240, AT3G15500, AT1G77450, AT1G32770, AT1G32870,
AT5G18037, AT3G10490, AT3G17730, AT4G28530, AT3G04430, AT3G03200, AT1G01720,
AT1G02210, AT2G02450, AT5G22380, AT1G34190, AT5G18300, AT4G01540, AT1G52880,
AT4G35580, AT3G56560, AT3G12977, AT3G04420, AT1G33060, AT1G61110, AT1G26870,
AT2G18060, AT5G04410, AT5G17260, AT5G53950, AT1G60340, AT1G62700, AT1G34180

269

Supplemental Table 4.7 (cont’d)
Plant zn
clust
SBP
SRF-TF

SWIM
TCP

WRKY

YABBY
zf-B box

zf-C2H2

zf-C2H2 4

AT2G23320, AT5G28650, AT3G04670, AT2G30590, AT2G24570, AT4G31550, AT4G24240
AT1G76580, AT1G69170, AT5G50670, AT1G02065, AT3G60030, AT2G42200, AT1G53160,
AT5G50570, AT1G20980, AT5G18830, AT3G57920, AT2G33810, AT2G47070, AT5G43270,
AT1G27370, AT1G27360, AT3G15270
AT1G31140, AT4G22950, AT5G15800, AT5G60910, AT3G66656, AT1G69120, AT2G26320,
AT5G39750, AT5G60440, AT3G04100, AT1G65360, AT5G49420, AT1G65300, AT2G24840,
AT1G26310, AT4G37940, AT5G27130, AT5G62165, AT2G14210, AT1G47760, AT3G05860,
AT1G60040, AT5G27580, AT5G20240, AT2G22540, AT4G36590, AT1G65330, AT5G49490,
AT5G58890, AT4G02235, AT2G34440, AT2G45650, AT5G26630, AT2G03060, AT3G57390,
AT1G77080, AT5G10140, AT1G77950, AT1G72350, AT1G31630, AT2G22630, AT5G51870,
AT5G48670, AT3G54340, AT1G28450, AT5G27810, AT2G03710, AT3G61120, AT1G22130,
AT1G22590, AT5G51860, AT3G57230, AT4G24540, AT5G65080, AT1G46408, AT5G40120,
AT3G02310, AT5G41200, AT5G38740, AT5G39810, AT5G65060, AT5G26650, AT5G27944,
AT5G38620, AT5G27090, AT5G04640, AT1G60920, AT5G06500, AT4G09960, AT5G40220,
AT5G65070, AT5G27070, AT1G01530, AT2G45660, AT5G26950, AT3G18650, AT1G77980,
AT1G17310, AT1G48150, AT5G26880, AT2G40210, AT5G37415, AT5G65050, AT1G28460,
AT1G59810, AT5G27960, AT1G69540, AT1G60880, AT5G55690, AT1G24260, AT4G18960,
AT2G42830, AT2G28700, AT4G11880, AT1G29962, AT5G23260, AT5G27050, AT4G11250,
AT1G18750, AT3G58780, AT5G26580, AT1G71692, AT5G65330, AT5G13790, AT3G30260,
AT1G31640
AT5G18960, AT5G28530, AT2G32250, AT4G19990, AT1G10240, AT1G76320, AT4G15090,
AT2G27110, AT4G38180, AT3G06250, AT1G80010
AT1G53230, AT5G51910, AT1G35560, AT5G08070, AT1G30210, AT1G72010, AT3G45150,
AT2G37000, AT5G41030, AT3G27010, AT3G02150, AT1G58100, AT1G68800, AT5G23280,
AT3G47620, AT1G67260, AT2G31070, AT1G69690, AT5G08330, AT3G18550, AT4G18390,
AT3G15030, AT2G45680, AT5G60970
AT5G45260, AT5G49520, AT1G30650, AT1G29280, AT1G80840, AT2G38470, AT2G30250,
AT1G64000, AT2G34830, AT4G04450, AT3G58710, AT4G23550, AT2G47260, AT2G46130,
AT5G46350, AT2G04880, AT5G56270, AT4G26440, AT2G37260, AT1G13960, AT4G12020,
AT1G69310, AT2G40740, AT5G24110, AT2G30590, AT4G01720, AT5G01900, AT5G64810,
AT1G29860, AT5G52830, AT1G80590, AT3G04670, AT5G07100, AT1G66550, AT4G31550,
AT2G46400, AT1G62300, AT2G23320, AT2G03340, AT2G21900, AT2G44745, AT3G56400,
AT1G18860, AT5G22570, AT4G39410, AT3G62340, AT5G41570, AT5G13080, AT3G01080,
AT1G69810, AT4G30935, AT4G22070, AT4G11070, AT4G01250, AT1G68150, AT4G24240,
AT1G66600, AT4G18170, AT5G28650, AT4G31800, AT2G24570, AT2G25000, AT1G55600,
AT5G45050, AT4G23810, AT5G43290, AT4G26640, AT5G26170, AT2G40750, AT3G01970,
AT5G15130, AT1G66560
AT4G00180, AT2G45190, AT1G69180, AT1G23420, AT2G26580, AT1G08465
AT5G57660, AT1G25440, AT4G10240, AT1G28050, AT5G15840, AT4G38960, AT2G31380,
AT1G78600, AT1G06040, AT3G07650, AT4G39070, AT1G75540, AT2G47890, AT5G15850,
AT4G15250, AT2G21320, AT2G24790, AT1G73870, AT2G33500, AT1G68520, AT5G48250,
AT5G24930, AT3G02380
AT3G10470, AT1G14580, AT1G02030, AT2G41835, AT3G48430, AT4G06634, AT5G66730,
AT2G45120, AT5G04390, AT1G26590, AT2G02070, AT3G45260, AT5G61470, AT5G03150,
AT1G72050, AT3G57670, AT2G17180, AT1G55110, AT3G60580, AT5G56200, AT1G68130,
AT2G02080, AT5G14140
AT5G61470, AT3G48430, AT3G20880, AT3G60580, AT1G08290, AT1G02030, AT1G51220,
AT1G13290, AT5G14140, AT1G72050, AT1G34790, AT2G45120, AT1G26590, AT4G06634

270

Supplemental Table 4.7 (cont’d)
zf-C2H2 6

zf-C2H2 jaz
zf-CCCH

zf-Dof

ZF-HD
dimer
zf-met

AT3G10470, AT4G16610, AT5G15480, AT5G59820, AT4G17810, AT5G67450, AT3G23130,
AT1G02030, AT5G01860, AT3G46080, AT3G53600, AT1G27730, AT2G28200, AT3G49930,
AT2G42410, AT2G45120, AT1G26590, AT4G35280, AT3G19580, AT5G61470, AT3G46070,
AT2G37740, AT3G46090, AT2G17180, AT5G03510, AT3G29340, AT3G60580, AT1G26610,
AT2G37430, AT5G04390, AT5G43170, AT5G56200, AT2G28710, AT5G04340
AT5G61470, AT1G27730, AT1G74250, AT2G28200, AT5G56200, AT5G26749, AT2G45120,
AT1G02030, AT2G26940, AT2G37430, AT1G26590, AT2G17180, AT3G60580
AT5G06770, AT3G12130, AT5G49200, AT3G12680, AT5G44260, AT3G19360, AT1G29560,
AT1G48195, AT5G06420, AT5G16540, AT1G32360, AT2G28450, AT5G58620, AT3G06410,
AT1G04990, AT2G47850, AT3G21810, AT1G29600, AT3G02830, AT2G32930, AT5G18550,
AT3G08505, AT1G10320, AT1G29570, AT1G68200, AT1G01350, AT5G63260, AT2G35430,
AT1G66810, AT1G75340, AT3G44785, AT3G48440, AT2G25900, AT1G03790
AT2G37590, AT1G51700, AT2G46590, AT5G60850, AT1G47655, AT3G47500, AT3G61850,
AT1G64620, AT1G29160, AT3G21270, AT1G28310, AT5G65590, AT1G26790, AT4G38000,
AT5G02460, AT2G34140, AT2G28510, AT4G21050, AT5G66940, AT3G45610, AT1G07640,
AT4G00940, AT4G21030, AT4G24060, AT3G52440, AT3G50410, AT3G55370, AT5G62940,
AT5G39660, AT4G21080, AT1G69570, AT1G21340, AT5G60200, AT4G21040, AT5G62430,
AT2G28810
AT3G50890, AT5G60480, AT5G15210, AT2G02540, AT1G69600, AT1G14687, AT4G24660,
AT2G18350, AT3G28920, AT1G18835, AT5G42780, AT1G74660, AT3G28917, AT1G75240,
AT5G65410, AT1G14440, AT5G39760
AT5G61470, AT1G27730, AT1G74250, AT1G02030

271

Supplemental Table 4.8 Best fit parameters of ODE models of the evolution of TF
expression above or below the ancestral state
Number of

Data

x

y

z

w

mu1

Sigma2

-2
log(L)3

Parameters
1

Control

0.089

0.089

0.089

0.089

0.002

0.002

-90.6

1

LightDev

0.095

0.095

0.095

0.095

0.004

0.003

-80.2

1

StressDiff

0.081

0.081

0.081

0.081

0.001

0.001

-100.7

1

Treatment

0.090

0.090

0.090

0.090

0.001

0.001

-94.0

2

Control

0.113

0.113

0.067

0.067

0.001

0.001

-98.0

2

LightDev

0.148

0.148

0.056

0.056

0.001

0.002

-125.0

2

StressDiff

0.074

0.074

0.087

0.087

0.001

0.001

-101.3

2

Treatment

0.121

0.121

0.064

0.064

0.001

0.001

-103.0

1. Mean of error distribution
2. Standard deviation of error distribution
3. -2 log-likelihood of model fit

272

Supplemental Table 4.9 Best fit parameters of ODE models of partitioning of ancestral
states between duplicate TFs
Number of

Data

x

y

z

w

mu1

sigma2

-2 log
L3

Parameters
1

Control

0.537

0.537

0.537

0.537

0.010

0.010

-75.5

1

LightDev

0.5298

0.5298

0.5298

0.5298

0.010

0.010

-75.3

1

StressDiff

0.471

0.471

0.471

0.471

0.008

0.008

-81.4

1

Treatment

0.054

0.054

0.054

0.05

0.010

0.010

-77.5

1

DapSeq

1.756

1.756

1.756

1.756

0.041

0.048

-38.8

2

Control

0.957

0.957

0.0779

0.0779

0.001

0.001

-140.3

2

LightDev

0.978

0.978

0.08

0.08

0.001

0.001

-139.4

2

StressDiff

0.747

0.747

0.0669

0.0669

0.001

0.001

-140.2

2

Treatment

1.368

1.368

0.0749

0.0749

1.70E-

0.001

-137.7

05
2

DapSeq

40.43

40.43

0.222

0.222

0.031

0.031

-48.8

4

DapSeq

10.12

4.1

0.2063

0.4175

0.003

0.004

-97.8

1. Mean of error distribution
2. Standard deviation of error distribution
3. -2 log-likelihood of model fit

273

Supplemental Table 4.10 The importance of all features used in the classification of
individual duplicate genes
Feature

Genome1

Kinases1

TFs1

Maximum Percent Identity (Paralog)

171.6 (1)

57.9 (1)

29.4 (2)

Sequence Conservation (Viridiplantae)

109.3 (2)

27.2 (2)

31.8 (1)

Gene Family Size (OrthoMCL)

81.6 (3)

2.0 (19)

27.7 (3)

Protein Length (in Amino Acids)

52.5 (4)

14.2 (3)

10.5 (11)

Expression Breadth (AtGenExpress)

46.2 (5)

9.4 (10)

11.2 (8)

Expression MAD/Median

41.0 (6)

5.4 (11)

11.8 (5)

Expression Mean (LightDev Data)

40.8 (7)

10.4 (7)

10.7 (9)

Expression Breadth (RNASeq)

40.0 (8)

3.3 (15)

11.5 (6)

Expression Mean (Control Data)

39.2 (9)

10.7 (6)

10.7 (10)

Expression Median (AtGenExpress)

37.5 (10)

10.2 (8)

10.2 (12)

Expression Mean (Stress Data)

37.0 (11)

11.5 (4)

12.0 (4)

Expression Mean (AtGenExpress)

36.9 (12)

11.3 (5)

11.3 (7)

Expression Median (RNASeq)

34.8 (13)

4.7 (12)

4.9 (16)

Sequence Conservation (Metazoa)

34.5 (14)

3.6 (13)

4.4 (18)

Nucleotide Diversity (Pi)

32.4 (15)

9.7 (9)

6.2 (14)

Expression Mean (Diff Data)

31.6 (16)

1.7 (2)

7.9 (13)

Sequence Conservation (Fungi)

30.6 (17)

2.4 (17)

4.6 (17)

Number of Protein Domains

28.4 (18)

2.3 (18)

5.6 (15)

Expression Mean (RNASeq)

18.6 (19)

3.0 (16)

2.9 (19)

Expression Maximum (RNASeq)

12.9 (20)

3.3 (14)

0.4 (20)

(AtGenExpress)

1: The importance of the feature as defined by the mean decrease in accuracy of the classification
when the feature is removed. Features are ordered according to the rank of their importance in
the whole genome model and the rank of each value for each model is indicated by ().

274

CHAPTER 5: CONCLUSION

275

In the preceding chapters, I have presented the application of mathematical modeling
techniques to my research, including the characterization of cyclic expression in two species and
the evolution of TF function and regulation. Here, I will further discuss the results of these
models as well as future directions for my research specifically and the application of modeling
to the understanding of gene expression in general.
PREDICTING CYCLIC EXPRESSION PATTERNS USING CIS-REGULATORY
ELEMENTS
In the work presented in Chapter 2, we identified genes that were cyclically expressed
under diel conditions in the algae C. reinhardtii and clustered them according their phase of
expression. Using these phase clusters, we found pCREs enriched in the promoters of genes
sharing the same phase of expression and trained a SVM model of diel expression using these
pCREs as features. The resulting model was able to predict the expression phase of diel genes
better than both random guessing and naive classifiers, but performed worse than previous SVM
models for predicting stress response in A. thaliana (Zou et al., 2011). However, we were able to
improve the performance of our classifier for subsets of cyclic genes by subdividing phaseclusters according to enriched gene functions. This improvement was not unexpected, given the
strict association between annotated gene function and phase of expression observed in our
study, but still indicates that the identified pCREs are important for regulating a subset of diel
expressed genes.
Continuing to look at cyclic expression, Chapter 3 describes a project where we changed
the system to cell-cycle expression in S. cerevisiae in order to take advantage of the extensive
expression and regulatory data available in this species. Here, we found that classifiers built
using TF-target interactions derived from ChIP-Chip, TF Deletion, and PWM data performed

276

better at predicting the phase of cell-cycle expression than the pCRE model of C. reinhardtii diel
expression. This improved performance is in part dependent on the near complete coverage of
TFs in S. cerevisiae by each data set, as reducing coverage of the best performing data sets to
less than half of the total S. cerevisiae TFs resulted in models with similar performance to those
in C. reinhardtii. Conversely, performance was improved by using interactions between TFs that
were part of regulatory FFLs and by combining features from the ChIP-Chip and Deletion data
sets. Subsequent importance and network analysis of features revealed that both canonical cellcycle regulators and at least two modules of TFs which lack evidence of being cell-cycle
regulators are amongst the best predictors of cell-cycle expression.
Nevertheless, even in our final model of S. cerevisiae cell-cycle regulation, there is room
to improve classification and learn more about the regulatory mechanisms controlling
expression. Using the optimal scoring threshold, our best cell-cycle model had an ~5% false
positive rate across the different phase classes, but recovered only half of the known cell-cycle
genes in each class. Why are these genes improperly classified? What aspects of cell-cycle
regulation are missing from our model? In the case of cell-cycle expression, it is known that
cyclin-dependent kinases play a key role in controlling cell-cycle progression both through direct
regulation of cell-cycle proteins and indirectly by modulating TF activity (CsikĂĄsz-Nagy et al.,
2009). In particular, phosphorylation controls the activity of Swi6 (Sidorova et al., 1995;
Lomberk et al., 2006; Shimada et al., 2009), a known regulator of cell-cycle initiation (Ho et al.,
1999) and one of the most important features in our model of cell-cycle expression. Swi6 itself
participates in chromatin remodeling (Grewal and Elgin, 2002; Haldar et al., 2011), thus cyclindependent kinases may further, indirectly regulate TF activity by affecting DNA accessibility.
Likewise, posttranscriptional regulation is of transcription is thought to play a role in

277

maintenance of diel and circadian cycles (Kojima et al., 2011; Romanowski and Yanovsky,
2015). Altering the expression of RNA-binding proteins in C. reinhardtii can not only lead to the
loss of cyclic expression, but disrupt the normal phase of cycling genes (Iliev et al. 2006),
indicating the post-transcriptional regulation exerts fine-grain control of the timing of circadian
and diel cycles. These additional layers of regulation, protein activation, DNA accessibility and
post-transcriptional regulation, likely hold the key to improving models of cyclic expression
beyond what is able to be done with CREs and pCREs alone.
DUPLICATION AND EVOLUTION OF TRANSCRIPTION FACTORS
While our previous models focused on the relationship between existing TFs and their
target genes, in model described in Chapter 4 we looked at the retention of duplicate genes pairs
following WGD events and specifically inferred the changes in expression and regulation of TFs
post-duplication. We found that TFs duplicates are retained more often than expected compared
to genes with other general functions (i.e. kinases, transporter activity, defense response), owing
in part to the fact that sequence divergence between duplicate TFs is on average greater than that
of other duplicate genes. Looking further into the relationship between duplicate TFs, we found
that the loss of an ancestral expression pattern or CREs from one duplicate copy but not the other
occurred more frequently than expected by random chance. Furthermore, loss of ancestral
expression and CREs occurred asymmetrically, such that duplicated TF pairs could be divided
into distinct “ancestral” copy, which retains almost all ancestral expression and regulatory sites,
and “non-ancestral” copy, which losses almost all ancestral expression and cis-regulatory sites,
but gains novel cis-regulatory sites instead. Furthermore, the partitioning of ancestral states is not
random as loss of ancestral expression in the first duplicate copy occurs at approximately an
order of magnitude faster than in the second duplicate copy, and the loss of ancestral regulation

278

occurs two orders of magnitude faster. We theorize that the preference for partitioning retained
duplicate TFs in ancestral and non-ancestral copies is due to the neofunctionalization of the nonancestral copy. This hypothesis is supported by examples where there is experimental evidence
supporting neo-functionalization of the non-ancestral copy of a TF duplicate pair.
However, expression and regulation are, at best, only a proxy for biological function. Our
model of TF evolution is, therefore, restricted by the fact that we cannot characterize gain or loss
of specific functions in the same way we can identify changes in expression or cis-regulatory
sites. The appropriate way to define biological function has been the subject of some controversy
(ENCODE Project Consortium, 2012; Doolittle, 2013; Graur et al., 2013; Brunet and Doolittle,
2014; Kellis et al., 2014a, Kellis et al., 2014b), and while a definition of function relying of
biochemical activity is useful for the purpose of broad categorization (e.g. transcription factor,
kinase, etc.), when considering gene evolution and retention, the definition of biological function
requires evidence of selection as a criterion. Assessing selection on large scale is difficult, even
under laboratory conditions (Winzeler et al., 1999; Tong et al., 2001), so an extensive catalog of
selective function would be challenging to produce, particularly for organisms with comparably
larger genomes and longer life-cycles. Thus, modeling the evolution of functions in the same
way we modeled the evolution of expression and regulation may not be possible, but a different
type of biological model might offer us an alternative approach. Recent studies have shown that
essential (Lloyd et al., 2015) and functional (Tsai et al., 2017) regions of the genome can be
classified by their molecular, genetic, and evolutionary features, including expression,
conservation, and interactions. Classification is done by scoring regions/genes based on their
features and comparing that score to the distribution of scores for genes that are known to be
essential, genes with known functions, and putatively non-functional regions such as

279

pseudogenes. In this way, the likelihood that a gene is essential, functional, or non-functional can
be quantified. Therefore, the evolution of the functions of retained duplicates may be explored in
a probabilistic manner by comparing the so called “functional-likelihood” score of non-ancestral
duplicates across different groups of genes as well as to the functional likelihood of ancestral
duplicate copies.
FUTURE PROSPECTS FOR MODELING GENE EXPRESSION
Modeling gene expression, particularly complex patterns sensitive to variations in time,
location or condition, is poised to benefit greatly from the continued improvement of sequencing
technology and computational methods. Independent of any particular sequencing approach, the
declining cost of sequencing DNA (Wetterstrand 2016) means that quantifying expression under
broader sets of conditions or with greater detail have become progressively more feasible. For
circadian and diel expression in particular, the increase in sampling resolution has generally lead
to the identification of more genes having cyclic patterns of expression (Harmer et al., 2000;
Edwards et al., 2006; Michael et al., 2008; Panchy et al., 2014; Zones et al., 2015). Long read
sequencing technologies, such as PacBio, also hold the promise of improving how we quantify
expression by removing the need to infer expression from short-reads that represent a fragment
of the actual transcript. Currently, full-length transcript sequencing data from PacBio have been
used to reassemble and re-annotate the genome of Triticum aestivum (Liu et al., 2017) as well as
profile the transcriptomes of Schizosaccharomyces pombe (Kuang et al., 2017) and Oryctolagus
cuniculus (Chen et al., 2017; Liu et al., 2017). However, short-read Illumina sequences are
needed in conjunction with PacBio reads due to the high per-base error rate of this technology
(Chen et al., 2017; Liu et al., 2017). As PacBio continues to improve and/or other technologies
become available that can offer the same long-read sequencing with fewer errors we can expect

280

to see transcriptome annotation and transcription quantification done using full-length transcript
sequences alone. In addition to less costly, higher quality quantification of expression, advances
in computing offer new opportunities for expression modeling. Specifically, so called “deep
learning” approaches involving neural networks have provided solutions to problems where
learning approaches previously had failed (Silver et al., 2016; Litjens et al., 2017). Yet, like other
machine learning methods, the models created by deep learning are a “black box” when it comes
to deriving biological significance (Albrecht et al., 2017). Though these new technologies offer
great potential for both generating and analyzing expression data, it will be up to enterprising
modelers to find way to apply these new approaches to answer biological questions.

281

REFERENCES

282

REFERENCES

Albert FW, Kruglyak L (2015) The role of regulatory variation in complex traits and disease.
Nat Rev Genet 16: 197–212
Albert R, Othmer HG (2003) The topology of the regulatory interactions predicts the
expression pattern of the segment polarity genes in Drosophila melanogaster. J Theor Biol
223: 1–18
Albrecht T, Slabaugh G, Alonso E, Al-Arif SMMR (2017) Deep learning for single-molecule
science. Nanotechnology 28: 423001
Alipanahi B, Delong A, Weirauch MT, Frey BJ (2015) Predicting the sequence specificities of
DNA- and RNA-binding proteins by deep learning. Nat Biotechnol 33: 831–838
Alon U (2007) Network motifs: theory and experimental approaches. Nat Rev Genet 8: 450–461
Alon U (2007) An Introduction to Systems Biology. Chapman & Hall/CRC, London
Alvarez-Ponce D, Fare M (2012) Evolutionary rate and duplicability in the Arabidopsis
thaliana protein-protein interaction network. Genome Biol Evol 4: 1247–1263
Anders S, Huber W (2010) Differential expression analysis for sequence count data. Genome
Biol 11: R106
Archambault V, Li CX, Tackett AJ, Wasch R, Chait BT, Rout MP, Cross FR (2003)
Genetic and biochemical evaluation of the importance of Cdc6 in regulating mitotic exit.
Mol Biol Cell 14: 4592–4604
Arimura G, Kopke S, Kunert M, Volpe V, David A, Brand P, Dabrowska P, Maffei ME,
Boland W (2008) Effects of feeding Spodoptera littoralis on lima bean leaves: IV. Diurnal
and nocturnal damage differentially initiate plant volatile emission. Plant Physiol 146: 965–
973
Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM, Davis AP, Dolinski K,
Dwight SS, Eppig JT, et al (2000) Gene ontology: tool for the unification of biology. The
Gene Ontology Consortium. Nat Genet 25: 25–29
Ay A, Arnosti DN (2011) Mathematical modeling of gene expression: a guide for the perplexed
biologist. Crit Rev Biochem Mol Biol 46: 137–151
Badis G, Berger MF, Philippakis AA, Talukder S, Gehrke AR, Jaeger SA, Chan ET,
Metzler G, Vedenko A, Chen X, et al (2009) Diversity and complexity in DNA
recognition by transcription factors. Science 324: 1720–1723
283

Bahler J (2005) Cell-cycle control of gene expression in budding and fission yeast. Annu Rev
Genet 39: 69–94
Bailey TL, Boden M, Buske FA, Frith M, Grant CE, Clementi L, Ren J, Li WW, Noble WS
(2009) MEME SUITE: tools for motif discovery and searching. Nucleic Acids Res 37:
W202--8
Bailey TL, Williams N, Misleh C, Li WW (2006) MEME: discovering and analyzing DNA and
protein sequence motifs. Nucleic Acids Res 34: W369--73
Bailey T, Krajewski P, Ladunga I, Lefebvre C, Li Q, Liu T, Madrigal P, Taslim C, Zhang J
(2013) Practical guidelines for the comprehensive analysis of ChIP-seq data. PLoS Comput
Biol 9: e1003326
Baker CR, Hanson-Smith V, Johnson AD (2013) Following gene duplication, paralog
interference constrains transcriptional circuit evolution. Science 342: 104–108
Baldwin IT, Meldau S (2013) Just in time: circadian defense patterns and the optimal defense
hypothesis. Plant Signal Behav 8: e24410
Bansal M, Della Gatta G, di Bernardo D (2006) Inference of gene regulatory networks and
compound mode of action from time course gene expression profiles. Bioinformatics 22:
815–822
Bean JM, Siggia ED, Cross FR (2005) High functional overlap between MluI cell-cycle box
binding factor and Swi4/6 cell-cycle box binding factor in the G1/S transcriptional program
in Saccharomyces cerevisiae. Genetics 171: 49–61
Beer MA, Tavazoie S (2004) Predicting gene expression from sequence. Cell 117: 185–198
Benjamini Y, Hochberg Y (1995) Controlling the False Discovery Rate - a Practical and
Powerful Approach to Multiple Testing. J R Stat Soc Ser B-Methodological 57: 289–300
Benveniste D, Sonntag H-J, Sanguinetti G, Sproul D (2014) Transcription factor binding
predicts histone modifications in human cell lines. Proc Natl Acad Sci U S A 111: 13367–
13372
Berger MF, Bulyk ML (2009) Universal protein-binding microarrays for the comprehensive
characterization of the DNA-binding specificities of transcription factors. Nat Protoc 4:
393–411
Bergholdt R, Brorsson C, Lage K, Nielsen JH, Brunak S, Pociot F (2009) Expression
profiling of human genetic and protein interaction networks in type 1 diabetes. PLoS One 4:
e6250

284

Bertoli C, Skotheim JM, de Bruin RAM (2013) Control of cell cycle transcription during G1
and S phases. Nat Rev Mol Cell Biol 14: 518–528
Bintu L, Buchler NE, Garcia HG, Gerland U, Hwa T, Kondev J, Phillips R (2005)
Transcriptional regulation by the numbers: models. Curr Opin Genet Dev 15: 116–124
Birchler JA, Veitia RA (2007) The gene balance hypothesis: from classical genetics to modern
genomics. Plant Cell 19: 395–402
Birchler JA, Veitia RA (2010) The gene balance hypothesis: implications for gene regulation,
quantitative traits and evolution. New Phytol 186: 54–62
Bisova K, Krylov DM, Umen JG (2005) Genome-wide annotation and expression profiling of
cell cycle regulatory genes in Chlamydomonas reinhardtii. Plant Physiol 137: 475–491
Blanc G, Wolfe KH (2004) Functional divergence of duplicated genes formed by polyploidy
during Arabidopsis evolution. Plant Cell 16: 1679–1691
Blasing OE, Gibon Y, Gunther M, Hohne M, Morcuende R, Osuna D, Thimm O, Usadel B,
Scheible WR, Stitt M (2005) Sugars and circadian regulation make major contributions to
the global regulation of diurnal gene expression in Arabidopsis. Plant Cell 17: 3257–3281
Bolger AM, Lohse M, Usadel B (2014) Trimmomatic: a flexible trimmer for Illumina sequence
data. Bioinformatics 30: 2114–2120
Bowers JE, Chapman BA, Rong J, Paterson AH (2003) Unravelling angiosperm genome
evolution by phylogenetic analysis of chromosomal duplication events. Nature 422: 433–
438
Breeden L (2003) Periodic transcription: a cycle within a cycle. Curr Biol 13: 31–38
Bruce VG (1970) The biological clock in Chlamydomonas reinhardtii. J Protozool 17: 328–334
Brunet TDP, Doolittle WF (2014) Getting “function” right. Proc Natl Acad Sci U S A 111:
E3365
Buck MJ, Lieb JD (2004) ChIP-chip: considerations for the design, analysis, and application of
genome-wide chromatin immunoprecipitation experiments. Genomics 83: 349–360
Bullard JH, Purdom E, Hansen KD, Dudoit S (2010) Evaluation of statistical methods for
normalization and differential expression in mRNA-Seq experiments. BMC Bioinformatics.
11: 94
Bulyk ML (2007) Protein binding microarrays for the characterization of DNA-protein
interactions. Adv Biochem Eng Biotechnol 104: 65–85

285

Bustin SA (2000) Absolute quantification of mRNA using real-time reverse transcription
polymerase chain reaction assays. J Mol Endocrinol 25: 169–193
Byrne TE, Wells MR, Johnson CH (1992) Circadian rhythms of chemotaxis to ammonium and
of methylammonium uptake in chlamydomonas. Plant Physiol 98: 879–886
Carretero-Paulet L, Fares MA (2012) Evolutionary dynamics and functional specialization of
plant paralogs formed by whole and small-scale genome duplications. Mol Biol Evol 29:
3541–3551
Cavalier-Smith T (1974) Basal body and flagellar development during the vegetative cell cycle
and the sexual cycle of Chlamydomonas reinhardii. J Cell Sci 16: 529–556
Chen K, Durand D, Farach-Colton M (2000) NOTUNG: a program for dating gene
duplications and optimizing gene family trees. J Comput Biol 7: 429–447
Chen KC, Calzone L, Csikasz-Nagy A, Cross FR, Novak B, Tyson JJ (2004) Integrative
analysis of cell cycle control in budding yeast. Mol Biol Cell 15: 3841–3862
Chen S-Y, Deng F, Jia X, Li C, Lai S-J (2017) A transcriptome atlas of rabbit revealed by
PacBio single-molecule long-read sequencing. Sci Rep 7: 7648
Chen YB, Dominic B, Mellon MT, Zehr JP (1998) Circadian rhythm of nitrogenase gene
expression in the diazotrophic filamentous nonheterocystous cyanobacterium
Trichodesmium sp. strain IMS 101. J Bacteriol 180: 3598–3605
Chen Y, Li Y, Narayan R, Subramanian A, Xie X (2016) Gene expression inference with
deep learning. Bioinformatics 32: 1832–1839
Chen Y, Elenee Argentinis JD, Weber G (2016) IBM Watson: How Cognitive Computing Can
Be Applied to Big Data Challenges in Life Sciences Research. Clin Ther 38: 688–701
Cheng C, Yan K-K, Yip KY, Rozowsky J, Alexander R, Shou C, Gerstein M (2011) A
statistical framework for modeling gene expression using chromatin features and
application to modENCODE datasets. Genome Biol 12: R15
Chikina MD, Huttenhower C, Murphy CT, Troyanskaya OG (2009) Global prediction of
tissue-specific gene expression and context-dependent gene networks in Caenorhabditis
elegans. PLoS Comput Biol 5: e1000417
Conesa A, Madrigal P, Tarazona S, Gomez-Cabrero D, Cervera A, McPherson A,
Szcześniak M, Gaffney DJ, Elo LL, Zhang X, et al (2016) A survey of best practices for
RNA-seq data analysis. Genome Biol 17: 13
Cooper-Knock J, Kirby J, Ferraiuolo L, Heath PR, Rattray M, Shaw PJ (2012) Gene
expression profiling in human neurodegenerative disease. Nat Rev Neurol 8: 518–530
286

Corellou F, Schwartz C, Motta JP, Djouani-Tahri el B, Sanchez F, Bouget FY (2009)
Clocks in the green lineage: comparative functional analysis of the circadian architecture of
the picoeukaryote ostreococcus. Plant Cell 21: 3436–3449
CsikĂĄsz-Nagy A, Kapuy O, TĂłth A, PĂĄl C, Jensen LJ, Uhlmann F, Tyson JJ, NovĂĄk B
(2009) Cell cycle regulation by feed-forward loops coupling transcription and
phosphorylation. Mol Syst Biol 5: 236
Danko CG, Hyland SL, Core LJ, Martins AL, Waters CT, Lee HW, Cheung VG, Kraus
WL, Lis JT, Siepel A (2015) Identification of active transcriptional regulatory elements
from GRO-seq data. Nat Methods 12: 433–438
Das MK, Dai H-K (2007) A survey of DNA motif finding algorithms. BMC Bioinformatics 8
Suppl 7: S21
Davies JP, Grossman AR (1994) Sequences controlling transcription of the Chlamydomonas
reinhardtii beta 2-tubulin gene after deflagellation and during the cell cycle. Mol Cell Biol
14: 5165–5174
Davis S, Meltzer PS (2007) GEOquery: a bridge between the Gene Expression Omnibus (GEO)
and BioConductor. Bioinformatics 23: 1846–1847
De Bodt S, Maere S, Van de Peer Y (2005) Genome duplication and the origin of angiosperms.
Trends Ecol Evol 20: 591–597
de Boer CG, Hughes TR (2012) YeTFaSCo: a database of evaluated yeast transcription factor
sequence specificities. Nucleic Acids Res 40: D169--79
Dehal P, Boore JL (2005) Two rounds of whole genome duplication in the ancestral vertebrate.
PLoS Biol 3: e314
Doherty CJ, Kay SA (2010) Circadian control of global gene expression patterns. Annu Rev
Genet 44: 419–444
Dong S, Adams KL (2011) Differential contributions to the transcriptome of duplicated genes in
response to abiotic stresses in natural and synthetic polyploids. New Phytol 190: 1045–1057
Dong X, Greven MC, Kundaje A, Djebali S, Brown JB, Cheng C, Gingeras TR, Gerstein
M, GuigĂł R, Birney E, et al (2012) Modeling gene expression using chromatin features in
various cellular contexts. Genome Biol 13: R53
Doolittle WF (2013) Is junk DNA bunk? A critique of ENCODE. Proc Natl Acad Sci U S A
110: 5294–5300

287

Edwards KD, Anderson PE, Hall A, Salathia NS, Locke JCW, Lynn JR, Straume M, Smith
JQ, Millar AJ (2006) FLOWERING LOCUS C mediates natural variation in the hightemperature response of the Arabidopsis circadian clock. Plant Cell 18: 639–650
Eldar A, Dorfman R, Weiss D, Ashe H, Shilo B-Z, Barkai N (2002) Robustness of the BMP
morphogen gradient in Drosophila embryonic patterning. Nature 419: 304–308
ENCODE Project Consortium (2012) An integrated encyclopedia of DNA elements in the
human genome. Nature 489: 57–74
Faith JJ, Hayete B, Thaden JT, Mogno I, Wierzbowski J, Cottarel G, Kasif S, Collins JJ,
Gardner TS (2007) Large-scale mapping and validation of Escherichia coli transcriptional
regulation from a compendium of expression profiles. PLoS Biol 5: e8
Fang SC, de los Reyes C, Umen JG (2006) Cell size checkpoint control by the retinoblastoma
tumor suppressor pathway. PLoS Genet 2: e167
Farre EM (2012) The regulation of plant growth by the circadian clock. Plant Biol 14: 401–410
Fay M (2010) Two-sided Exact Tests and Matching Confidence Intervals for Discrete Data. R J
2: 53–58
Felsenstein J (2005) PHYLIP (Phylogeny Inference Package) version 3.6. . Distrib. by author.
Felsenstein J (1989) PHYLIP - Phylogeny Inference Package (Version 3.2). Cladistics 5: 164–
166
Femino AM, Fay FS, Fogarty K, Singer RH (1998) Visualization of single RNA transcripts in
situ. Science (80- ) 280: 585–590
Feng J, Liu T, Qin B, Zhang Y, Liu XS (2012) Identifying ChIP-seq enrichment using MACS.
Nat Protoc 7: 1728–1740
Fernyhough P (2001) Quantification of mRNA levels using northern blotting. Methods Mol
Biol 169: 53–63
Filichkin SA, Breton G, Priest HD, Dharmawardhana P, Jaiswal P, Fox SE, Michael TP,
Chory J, Kay SA, Mockler TC (2011) Global profiling of rice and poplar transcriptomes
highlights key conserved circadian-controlled pathways and cis-regulatory modules. PLoS
One 6: e16907
Finn RD, Coggill P, Eberhardt RY, Eddy SR, Mistry J, Mitchell AL, Potter SC, Punta M,
Qureshi M, Sangrador-Vegas A, et al (2016) The Pfam protein families database: towards
a more sustainable future. Nucleic Acids Res 44: D279--85

288

Fonken LK, Nelson RJ (2014) The Effects of Light at Night on Circadian Clocks and
Metabolism. Endocr Rev 35: 648-670.
Force A, Lynch M, Pickett FB, Amores A, Yan YL, Postlethwait J (1999) Preservation of
duplicate genes by complementary, degenerative mutations. Genetics 151: 1531–1545
Friedman N, Linial M, Nachman I, Pe’er D (2000) Using Bayesian networks to analyze
expression data. J Comput Biol 7: 601–620
Frund J, Dormann CF, Tscharntke T (2011) Linne’s floral clock is slow without pollinators-flower closure and plant-pollinator interaction webs. Ecol Lett 14: 896–904
Furey TS (2012) ChIP-seq and beyond: new and improved methodologies to detect and
characterize protein-DNA interactions. Nat Rev Genet 13: 840–852
Futhcher B (2002) Transcriptional regulatory networks and the yeast cell cycle. Curr Opin Cell
Biol 14: 676–683
Geeven G, van Kesteren RE, Smit AB, de Gunst MCM (2012) Identification of contextspecific gene regulatory networks with GEMULA--gene expression modeling using LAsso.
Bioinformatics 28: 214–221
Gene Ontology Consortium (2015) Gene Ontology Consortium: going forward. Nucleic Acids
Res 43: D1049--56
Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, Ellis B, Gautier L,
Ge Y, Gentry J, et al (2004) Bioconductor: open software development for computational
biology and bioinformatics. Genome Biol 5: R80
Glass L, Kauffman SA (1973) The logical analysis of continuous, non-linear biochemical
control networks. J Theor Biol 39: 103–129
Goda H, Sasaki E, Akiyama K, Maruyama-Nakashita A, Nakabayashi K, Li W, Ogawa M,
Yamauchi Y, Preston J, Aoki K, et al (2008) The AtGenExpress hormone and chemical
treatment data set: experimental design, data evaluation, model data analysis and data
access. Plant J 55: 526–542
Goodspeed D, Chehab EW, Min-Venditti A, Braam J, Covington MF (2012) Arabidopsis
synchronizes jasmonate-mediated defense with insect circadian behavior. Proc Natl Acad
Sci U S A 109: 4674–4677
Goodwin S, McPherson JD, McCombie WR (2016) Coming of age: ten years of nextgeneration sequencing technologies. Nat Rev Genet 17: 333–351

289

Graur D, Zheng Y, Price N, Azevedo RBR, Zufall RA, Elhaik E (2013) On the immortality
of television sets: “function” in the human genome according to the evolution-free gospel of
ENCODE. Genome Biol Evol 5: 578–590
Grewal SIS, Elgin SCR (2002) Heterochromatin: new possibilities for the inheritance of
structure. Curr Opin Genet Dev 12: 178–187
Grossman AR, Croft M, Gladyshev VN, Merchant SS, Posewitz MC, Prochnik S, Spalding
MH (2007) Novel metabolism in Chlamydomonas through the lens of genomics. Curr Opin
Plant Biol 10: 190–198
Gualberti G, Papi M, Bellucci L, Ricci I, Bouchez D, Camilleri C, Costantino P, Vittorioso
P (2002) Mutations in the Dof zinc finger genes DAG2 and DAG1 influence with opposite
effects the germination of Arabidopsis seeds. Plant Cell 14: 1253–1263
Guidi M, Ruault M, Marbouty M, Loiodice I, Cournac A, Billaudeau C, Hocher A,
Mozziconacci J, Koszul R, Taddei A (2015) Spatial reorganization of telomeres in longlived quiescent cells. Genome Biol 16: 206
Haldar S, Saini A, Nanda JS, Saini S, Singh J (2011) Role of Swi6/HP1 self-associationmediated recruitment of Clr4/Suv39 in establishment and maintenance of heterochromatin
in fission yeast. J Biol Chem 286: 9308–9320
Hall M, Frank E, Holmes G, Pfahringer B, Reutemann P, Witten IH (2009) The WEKA data
mining software: an update. SIGKDD Explor Newsl 11: 10–18
Hanada K, Zou C, Lehti-Shiu MD, Shinozaki K, Shiu S-H (2008) Importance of lineagespecific expansion of plant tandem duplicates in the adaptive response to environmental
stimuli. Plant Physiol 148: 993–1003
Harbison CT, Gordon DB, Lee TI, Rinaldi NJ, Macisaac KD, Danford TW, Hannett NM,
Tagne J-B, Reynolds DB, Yoo J, et al (2004) Transcriptional regulatory code of a
eukaryotic genome. Nature 431: 99–104
Harmer SL, Hogenesch JB, Straume M, Chang HS, Han B, Zhu T, Wang X, Kreps JA,
Kay SA (2000) Orchestrated transcription of key pathways in Arabidopsis by the circadian
clock. Science 290: 2110–2113
Harmer SL, Kay SA (2005) Positive and negative factors confer phase-specific circadian
regulation of transcription in Arabidopsis. Plant Cell 17: 1926–1940
Harris EH (2001) Chlamydomonas as a model organism. Annu Rev Plant Physiol Plant Mol
Biol 52: 363–406
Henriksen PA, Kotelevtsev Y (2002) Application of gene expression profiling to cardiovascular
disease. Cardiovasc Res 54: 16–24
290

Ho Y, Costanzo M, Moore L, Kobayashi R, Andrews BJ (1999) Regulation of transcription at
the Saccharomyces cerevisiae start transition by Stb1, a Swi6-binding protein. Mol Cell
Biol 19: 5267–5278
Hoffman MM, Buske OJ, Wang J, Weng Z, Bilmes JA, Noble WS (2012) Unsupervised
pattern discovery in human chromatin structure through genomic segmentation. Nat
Methods 9: 473–476
Hoheisel JD (2006) Microarray technology: beyond transcript profiling and genotype analysis.
Nat Rev Genet 7: 200–210
Hong T, Watanabe K, Ta CH, Villarreal-Ponce A, Nie Q, Dai X (2015) An Ovol2-Zeb1
Mutual Inhibitory Circuit Governs Bidirectional and Multi-step Transition between
Epithelial and Mesenchymal States. PLoS Comput Biol 11: e1004569
Honkela A, Girardot C, Gustafson EH, Liu Y-H, Furlong EEM, Lawrence ND, Rattray M
(2010) Model-based method for transcription factor target identification with limited data.
Proc Natl Acad Sci U S A 107: 7793–7798
Hu Q, Sommerfeld M, Jarvis E, Ghirardi M, Posewitz M, Seibert M, Darzins A (2008)
Microalgal triacylglycerols as feedstocks for biofuel production: perspectives and advances.
Plant J 54: 621–639
Hughes JD, Estep PW, Tavazoie S, Church GM (2000) Computational identification of cisregulatory elements associated with groups of functionally related genes in Saccharomyces
cerevisiae. J Mol Biol 296: 1205–1214
Hughes ME, Hogenesch JB, Kornacker K (2010) JTK_CYCLE: an efficient nonparametric
algorithm for detecting rhythmic components in genome-scale data sets. J Biol Rhythm 25:
372–380
Hwang S, Herrin DL (1994) Control of lhc gene transcription by the circadian clock in
Chlamydomonas reinhardtii. Plant Mol Biol 26: 557–569
Iliev D, Voytsekh O, Schmidt EM, Fiedler M, Nykytenko A, Mittag M (2006) A heteromeric
RNA-binding protein is involved in maintaining acrophase and period of the circadian
clock. Plant Physiol 142: 797–806
Irizarry R, Hobbs B, Collin F, Beazer-Barclay Y, Antonellis K, Scherf U, Speed T (2003)
Exploration, normalization, and summaries of high density oligonucleotide array probe
level data. Biostatistics 4: 249–264
Iyer VR, Horak CE, Scafe CS, Botstein D, Snyder M, Brown PO (2001) Genomic binding
sites of the yeast cell-cycle transcription factors SBF and MBF. Nature 409: 533–538

291

Jacobshagen S, Johnson CH (1994) Circadian rhythms of gene expression in Chlamydomonas
reinhardtii: circadian cycling of mRNA abundances of cab II, and possibly of beta-tubulin
and cytochrome c. Eur J Cell Biol 64: 142–152
Jaeger J, Blagov M, Kosman D, Kozlov KN, Manu, Myasnikova E, Surkova S, VanarioAlonso CE, Samsonova M, Sharp DH, et al (2004) Dynamical analysis of regulatory
interactions in the gap gene system of Drosophila melanogaster. Genetics 167: 1721–1737
Jarolim S, Ayer A, Pillay B, Gee AC, Phrakaysone A, Perrone GG, Breitenbach M, Dawes
IW (2013) Saccharomyces cerevisiae genes involved in survival of heat shock. G3 3: 2321–
2333
Jaskowiak PA, Campello RJGB, Costa IG (2014) On the selection of appropriate distances for
gene expression data clustering. BMC Bioinformatics 15 Suppl 2: S2
Jiang W-K, Liu Y-L, Xia E-H, Gao L-Z (2013) Prevalent role of gene features in determining
evolutionary fates of whole-genome duplication duplicated genes in flowering plants. Plant
Physiol 161: 1844–1861
Jiao Y, Wickett NJ, Ayyampalayam S, Chanderbali AS, Landherr L, Ralph PE, Tomsho
LP, Hu Y, Liang H, Soltis PS, et al (2011) Ancestral polyploidy in seed plants and
angiosperms. Nature 473: 97–100
Jimenez-Gomez JM, Corwin JA, Joseph B, Maloof JN, Kliebenstein DJ (2011) Genomic
analysis of QTLs and genes altering natural variation in stochastic noise. PLoS Genet 7:
e1002295
JimĂŠnez-GĂłmez JM, Wallace AD, Maloof JN (2010) Network analysis identifies ELF3 as a
QTL for the shade avoidance response in Arabidopsis. PLoS Genet 6: e1001100
Jin J, Zhang H, Kong L, Gao G, Luo J (2014) PlantTFDB 3.0: a portal for the functional and
evolutionary study of plant transcription factors. Nucleic Acids Res 42: D1182--7
Jolma A, Yin Y, Nitta KR, Dave K, Popov A, Taipale M, Enge M, Kivioja T, Morgunova E,
Taipale J (2015) DNA-dependent formation of transcription factor pairs alters their binding
specificity. Nature 527: 384–388
Jones RF (1970) Physiological and Biochemical Aspects of Growth and Gametogenesis in
Chlamydomonas-Reinhardtii. Ann N Y Acad Sci 175: 648
Juven-Gershon T, Hsu J-Y, Theisen JW, Kadonaga JT (2008) The RNA polymerase II core
promoter - the gateway to transcription. Curr Opin Cell Biol 20: 253–259
Karlebach G, Shamir R (2008) Modelling and analysis of gene regulatory networks. Nat Rev
Mol Cell Biol 9: 770–780

292

Katoh K, Misawa K, Kuma K, Miyata T (2002) MAFFT: a novel method for rapid multiple
sequence alignment based on fast Fourier transform. Nucleic Acids Res 30: 3059–3066
Katoh K, Standley DM (2013) MAFFT multiple sequence alignment software version 7:
improvements in performance and usability. Mol Biol Evol 30: 772–780
Kauffman S, Peterson C, Samuelsson B, Troein C (2003) Random Boolean network models
and the yeast transcriptional network. Proc Natl Acad Sci U S A 100: 14796–14799
Kazemian M, Pham H, Wolfe SA, Brodsky MH, Sinha S (2013) Widespread evidence of
cooperative DNA binding by transcription factors in Drosophila development. Nucleic
Acids Res 41: 8237–8252
Kellis M, Birren BW, Lander ES (2004) Proof and evolutionary analysis of ancient genome
duplication in the yeast Saccharomyces cerevisiae. Nature 428: 617–624
Kellis M, Wold B, Snyder MP, Bernstein BE, Kundaje A, Marinov GK, Ward LD, Birney
E, Crawford GE, Dekker J, et al (2014) Defining functional DNA elements in the human
genome. Proc Natl Acad Sci U S A 111: 6131–6138
Kellis M, Wold B, Snyder MP, Bernstein BE, Kundaje A, Marinov GK, Ward LD, Birney
E, Crawford GE, Dekker J, et al (2014) Reply to Brunet and Doolittle: Both selected
effect and causal role elements can influence human biology and disease. Proc Natl Acad
Sci U S A 111: E3366
Keng T (1992) HAP1 and ROX1 form a regulatory pathway in the repression of HEM13
transcription in Saccharomyces cerevisiae. Mol Cell Biol 12: 2616–2623
Kerr G, Ruskin HJ, Crane M, Doolan P (2008) Techniques for clustering gene expression
data. Comput Biol Med 38: 283–293
Kilian J, Whitehead D, Horak J, Wanke D, Weinl S, Batistic O, D’Angelo C, BornbergBauer E, Kudla J, Harter K (2007) The AtGenExpress global stress expression data set:
protocols, evaluation and model data analysis of UV-B light, drought and cold stress
responses. Plant J 50: 347–363
Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg SL (2013) TopHat2: accurate
alignment of transcriptomes in the presence of insertions, deletions and gene fusions.
Genome Biol 14: R36
Kinmonth-Schultz HA, Golembeski GS, Imaizumi T (2013) Circadian clock-regulated
physiological outputs: dynamic responses in nature. Semin Cell Dev Biol 24: 407–413
Kojima S, Shingle DL, Green CB (2011) Post-transcriptional control of circadian rhythms. J
Cell Sci 124: 311–320

293

Koranda M, Schleiffer A, Endler L, Ammerer G (2000) Forkhead-like transcription factors
recruit Ndd1 to the chromatin of G2/M-specific promoters. Nature 406: 94–98
Kuang Z, Boeke JD, Canzar S (2017) The dynamic landscape of fission yeast meiosis
alternative-splice isoforms. Genome Res 27: 145–156
Kucho K, Okamoto K, Tabata S, Fukuzawa H, Ishiura M (2005) Identification of novel
clock-controlled genes by cDNA macroarray analysis in Chlamydomonas reinhardtii. Plant
Mol Biol 57: 889–906
Kummerfeld SK, Teichmann SA (2006) DBD: a transcription factor prediction database.
Nucleic Acids Res 34: D74--81
Kwasnieski JC, Mogno I, Myers CA, Corbo JC, Cohen BA (2012) Complex effects of
nucleotide variants in a mammalian cis-regulatory element. Proc Natl Acad Sci U S A 109:
19498–19503
Landt SG, Marinov GK, Kundaje A, Kheradpour P, Pauli F, Batzoglou S, Bernstein BE,
Bickel P, Brown JB, Cayting P, et al (2012) ChIP-seq guidelines and practices of the
ENCODE and modENCODE consortia. Genome Res 22: 1813–1831
Laporte D, Courtout F, Tollis S, Sagot I (2016) Quiescent Saccharomyces cerevisiae forms
telomere hyperclusters at the nuclear membrane vicinity through a multifaceted mechanism
involving Esc1, the Sir complex, and chromatin condensation. Mol Biol Cell 27: 1875–1884
Lee TI, Young RA (2000) Transcription of eukaryotic protein-coding genes. Annu Rev Genet
34: 77–137
Lee T-H, Tang H, Wang X, Paterson AH (2013) PGDD: a database of gene and genome
duplication in plants. Nucleic Acids Res 41: D1152--8
Lehmann M, Gustav D, Galizia CG (2011) The early bee catches the flower - circadian
rhythmicity influences learning performance in honey bees, Apis mellifera. Behav Ecol
Sociobiol 65: 205–215
Lehti-Shiu M, Panchy N, Wang P, Uygun S, Shiu S-H (2016) Diversity, expansion, and
evolutionary novelty of plant DNA-binding transcription factor families. BBA 1860: 3–20
Lelli KM, Slattery M, Mann RS (2012) Disentangling the many layers of eukaryotic
transcriptional regulation. Annu Rev Genet 46: 43–68
Lespinet O, Wolf YI, Koonin E V, Aravind L (2002) The role of lineage-specific gene family
expansion in the evolution of eukaryotes. Genome Res 12: 1048–1059
Li F, Long T, Lu Y, Ouyang Q, Tang C (2004) The yeast cell-cycle network is robustly
designed. Proc Natl Acad Sci U S A 101: 4781–4786
294

Li M, Hada A, Sen P, Olufemi L, Hall MA, Smith BY, Forth S, McKnight JN, Patel A,
Bowman GD, et al (2015) Dynamic regulation of transcription factors by nucleosome
remodeling. Elife 4
Li S, Brazhnik P, Sobral B, Tyson JJ (2008) A quantitative study of the division cycle of
Caulobacter crescentus stalked cells. PLoS Comput Biol 4: e9
Li Y, Chen C-Y, Kaye AM, Wasserman WW (2015) The identification of cis-regulatory
elements: A review from a machine learning perspective. Biosystems 138: 6–17
Li Y, Chen C-Y, Wasserman WW (2016) Deep Feature Selection: Theory and Application to
Identify Enhancers and Promoters. J Comput Biol 23: 322–336
Li Y, Liang M, Zhang Z (2014) Regression analysis of combined gene expression regulation in
acute myeloid leukemia. PLoS Comput Biol 10: e1003908
Li Z, Defoort J, Tasdighian S, Maere S, de Peer Y, De Smet R (2016) Gene Duplicability of
Core Genes Is Highly Consistent across All Angiosperms. Plant Cell 28: 326–344
Libbrecht MW, Noble WS (2015) Machine learning applications in genetics and genomics. Nat
Rev Genet 16: 321–332
Litjens G, Kooi T, Bejnordi BE, Setio AAA, Ciompi F, Ghafoorian M, van der Laak
JAWM, van Ginneken B, SĂĄnchez CI (2017) A survey on deep learning in medical image
analysis. Med Image Anal 42: 60–88
Litt A, Irish V (2003) Duplication and Diversification in the APETALA1/FRUITFULL Floral
Homeotic Gene Lineage: Implications for the Evolution of Floral Development. Genetics
165: 821–833
Liu B, Hsu W, Ma Y (1998) Integrating Classification and Association Rule Mining. Proc.
Fourth Int. Conf. Knowl. Discov. Data Min.
Liu H, Smith TPL, Nonneman DJ, Dekkers JCM, Tuggle CK (2017) A high-quality
annotated transcriptome of swine peripheral blood. BMC Genomics 18: 479
Liu M-J, Seddon AE, Tsai ZT-Y, Major IT, Floer M, Howe GA, Shiu S-H (2015)
Determinants of nucleosome positioning and their influence on plant gene expression.
Genome Res 25: 1182–1195
Lloyd JP, Seddon AE, Moghe GD, Simenc MC, Shiu S-H (2015) Characteristics of Plant
Essential Genes Allow for within- and between-Species Prediction of Lethal Mutant
Phenotypes. Plant Cell 27: 2133–2147
Lomberk G, Bensi D, Fernandez-Zapico ME, Urrutia R (2006) Evidence for the existence of
an HP1-mediated subcode within the histone code. Nat Cell Biol 8: 407–415
295

Lynch M, Conery JS (2000) The evolutionary fate and consequences of duplicate genes.
Science (80- ) 290: 1151–1155
Lynch M, Conery JS (2003) The evolutionary demography of duplicate genes. J Struct Funct
Genomics 3: 35–44
Lyons E, Pedersen B, Kane J, Alam M, Ming R, Tang H, Wang X, Bowers J, Paterson A,
Lisch D, et al (2008) Finding and comparing syntenic regions among Arabidopsis and the
outgroups papaya, poplar, and grape: CoGe with rosids. Plant Physiol 148: 1772–1781
Macneil LT, Walhout AJM (2011) Gene regulatory networks and the role of robustness and
stochasticity in the control of gene expression. Genome Res 21: 645–657
Maere S, De Bodt S, Raes J, Casneuf T, Van Montagu M, Kuiper M, de Peer Y (2005)
Modeling gene and genome duplications in eukaryotes. Proc Natl Acad Sci U S A 102:
5454–5459
Marbach D, Costello JC, KĂźffner R, Vega NM, Prill RJ, Camacho DM, Allison KR,
DREAM5 Consortium, Kellis M, Collins JJ, et al (2012) Wisdom of crowds for robust
gene network inference. Nat Methods 9: 796–804
Marbach D, Prill RJ, Schaffter T, Mattiussi C, Floreano D, Stolovitzky G (2010) Revealing
strengths and weaknesses of methods for gene network inference. Proc Natl Acad Sci U S A
107: 6286–6291
Margolin AA, Nemenman I, Basso K, Wiggins C, Stolovitzky G, Dalla Favera R, Califano
A (2006) ARACNE: an algorithm for the reconstruction of gene regulatory networks in a
mammalian cellular context. BMC Bioinformatics 7 Suppl 1: S7
Matsuo T, Ishiura M (2010) New insights into the circadian clock in Chlamydomonas. Int Rev
Cell Mol Biol 280: 281–314
McCarthy E, Mohamed A, Litt A (2015) Functional Divergence of APETALA1 and
FRUITFULL is due to Changes in both Regulation and Coding Sequence. Front Plant Sci 6:
1076
McNabb SL, Truman JW (2008) Light and peptidergic eclosion hormone neurons stimulate a
rapid eclosion response that masks circadian emergence in Drosophila. J Exp Biol 211:
2263–2274
Michael TP, McClung CR (2002) Phase-specific circadian clock regulatory elements in
Arabidopsis. Plant Physiol 130: 627–638
Michael TP, Mockler TC, Breton G, McEntee C, Byer A, Trout JD, Hazen SP, Shen R,
Priest HD, Sullivan CM, et al (2008) Network discovery pipeline elucidates conserved
time-of-day-specific cis-regulatory modules. PLoS Genet 4: e14
296

Miller JA, Widom J (2003) Collaborative competition mechanism for gene activation in vivo.
Mol Cell Biol 23: 1623–1632
Min S, Lee B, Yoon S (2016) Deep learning in bioinformatics. Brief. Bioinform.
Mistry J, Finn RD, Eddy SR, Bateman A, Punta M (2013) Challenges in homology search:
HMMER3 and convergent evolution of coiled-coil regions. Nucleic Acids Res 41: e121
Mittag M, Kiaulehn S, Johnson CH (2005) The circadian clock in Chlamydomonas reinhardtii.
What is it for? What is it similar to? Plant Physiol 137: 399–409
Moghe GD, Hufnagel DE, Tang H, Xiao Y, Dworkin I, Town CD, Conner JK, Shiu S-H
(2014) Consequences of Whole-Genome Triplication as Revealed by Comparative Genomic
Analyses of the Wild Radish Raphanus raphanistrum and Three Other Brassicaceae Species.
Plant Cell 26: 1925–1937
Moghe GD, Shiu S-H (2014) The causes and molecular consequences of polyploidy in
flowering plants. Ann N Y Acad Sci 1320: 16–34
Monfared M, Simon M, Meister R, Roin-Villanova I, Kooiker M, Colombo L, Fletcher J,
Gasser C (2011) Overlapping and antagonistic activities of BASIC PENTACYSTEINE
genes affect a range of developmental processes in Arabidopsis. Plant J 66: 1020–1031
Monnier A, Liverani S, Bouvet R, Jesson B, Smith JQ, Mosser J, Corellou F, Bouget FY
(2010) Orchestrated transcription of biological processes in the marine picoeukaryote
Ostreococcus exposed to light/dark cycles. BMC Genomics 11: 192
Moreau H, Verhelst B, Couloux A, Derelle E, Rombauts S, Grimsley N, Van Bel M, Poulain
J, Katinka M, Hohmann-Marriott MF, et al (2012) Gene functionalities and genome
structure in Bathycoccus prasinos reflect cellular specializations at the base of the green
lineage. Genome Biol 13: R74
Myburg AA, Grattapaglia D, Tuskan GA, Hellsten U, Hayes RD, Grimwood J, Jenkins J,
Lindquist E, Tice H, Bauer D, et al (2014) The genome of Eucalyptus grandis. Nature
510: 356–362
Nachman I, Regev A, Friedman N (2004) Inferring quantitative models of regulatory networks
from expression data. Bioinformatics 20 Suppl 1: i248--56
Nica AC, Dermitzakis ET (2013) Expression quantitative trait loci: present and future. Philos
Trans R Soc Lond B Biol Sci 368: 20120362
Nolan T, Hands RE, Bustin SA (2006) Quantification of mRNA using real-time RT-PCR. Nat
Protoc 1: 1559–1582

297

Oakley T, Østman B, Wilson A (2006) Repression and loss of gene expression outpaces
activation and gain in recently duplicated fly genes. Proc Natl Acad Sci U S A 103: 11637–
11641
Ohno S (1970) Evolution by Gene Duplication. Springer-Verlag, New York
Olson BJ, Oberholzer M, Li Y, Zones JM, Kohli HS, Bisova K, Fang SC, Meisenhelder J,
Hunter T, Umen JG (2010) Regulation of the Chlamydomonas cell cycle by a stable,
chromatin-associated retinoblastoma tumor suppressor complex. Plant Cell 22: 3331–3347
O’Malley RC, Huang S-SC, Song L, Lewsey MG, Bartlett A, Nery JR, Galli M, Gallavotti
A, Ecker JR (2016) Cistrome and Epicistrome Features Shape the Regulatory DNA
Landscape. Cell 165: 1280–1292
Oyelade J, Isewon I, Oladipupo F, Aromolaran O, Uwoghiren E, Ameh F, Achas M,
Adebiyi E (2016) Clustering Algorithms: Their Application to Gene Expression Data.
Bioinform Biol Insights 10: 237–253
Pagel M, Meade A, Barker D (2004) Bayesian estimation of ancestral character states on
phylogenies. Syst Biol 53: 673–684
Pajuelo E, Pajuelo P, Clemente MT, Marquez AJ (1995) Regulation of the expression of
ferredoxin-nitrite reductase in synchronous cultures of Chlamydomonas reinhardtii.
Biochim Biophys Acta 1249: 72–78
Panchy N, Lehti-Shiu M, Shiu S-H (2016) Evolution of Gene Duplication in Plants. Plant
Physiol 171: 2294–2316
Panchy N, Wu G, Newton L, Tsai C-H, Chen J, Benning C, FarrĂŠ EM, Shiu S-H (2014)
Prevalence, evolution, and cis-regulation of diel transcription in Chlamydomonas
reinhardtii. G3 4: 2461–2471
Panda S, Antoch MP, Miller BH, Su AI, Schook AB, Straume M, Schultz PG, Kay SA,
Takahashi JS, Hogenesch JB (2002) Coordinated transcription of key pathways in the
mouse by the circadian clock. Cell 109: 307–320
Panopoulou G, Hennig S, Groth D, Krause A, Poustka AJ, Herwig R, Vingron M, Lehrach
H (2003) New evidence for genome-wide duplications at the origin of vertebrates using an
amphioxus gene set and completed animal genomes. Genome Res 13: 1056–1066
Pett JP, Korenčič A, Wesener F, Kramer A, Herzel H (2016) Feedback Loops of the
Mammalian Circadian Clock Constitute Repressilator. PLoS Comput Biol 12: e1005266
Pique-Regi R, Degner JF, Pai AA, Gaffney DJ, Gilad Y, Pritchard JK (2011) Accurate
inference of transcription factor binding from DNA sequence and chromatin accessibility
data. Genome Res 21: 447–455
298

Price C, Nasmyth K, Shuster T (1991) A general approach to the isolation of cell cycleregulated genes in the budding yeast, Saccharomyces cerevisiae. J Mol Biol 218: 543–556
Punta M, Coggill PC, Eberhardt RY, Mistry J, Tate J, Boursnell C, Pang N, Forslund K,
Ceric G, Clements J, et al (2012) The Pfam protein families database. Nucleic Acids Res
40: D290-301
Qian W, Liao B, Chang A, Zhang J (2011) Maintenance of duplicate genes and their functional
redundancy by reduced expression. Trends Genet 26: 425–430
Ral JP, Colleoni C, Wattebled F, Dauvillee D, Nempont C, Deschamps P, Li ZY, Morell
MK, Chibbar R, Purton S, et al (2006) Circadian clock regulation of starch metabolism
establishes GBSSI as a major contributor to amylopectin synthesis in Chlamydomonas
reinhardtii. Plant Physiol 142: 305–317
Rapaport F, Khanin R, Liang Y, Pirun M, Krek A, Zumbo P, Mason CE, Socci ND, Betel
D (2013) Comprehensive evaluation of differential gene expression analysis methods for
RNA-seq data. Genome Biol 14: R95
Ratcliff WC, Herron MD, Howell K, Pentz JT, Rosenzweig F, Travisano M (2013)
Experimental evolution of an alternating uni- and multicellular life cycle in
Chlamydomonas reinhardtii. Nat Commun 4: 2742
Reimand J, Vaquerizas JM, Todd AE, Vilo J, Luscombe NM (2010) Comprehensive
reanalysis of transcription factor knockout expression data in Saccharomyces cerevisiae
reveals many new targets. Nucleic Acids Res 38: 4768–4777
Renny-Byfield S, Gallagher JP, Grover CE, Szadkowski E, Page JT, Udall JA, Wang X,
Paterson AH, Wendel JF (2014) Ancient gene duplicates in Gossypium (cotton) exhibit
near-complete expression divergence. Genome Biol Evol 6: 559–571
Reuter JA, Spacek D V, Snyder MP (2015) High-throughput sequencing technologies. Mol
Cell 58: 586–597
Robinson MD, McCarthy DJ, Smyth GK (2010) edgeR: a Bioconductor package for
differential expression analysis of digital gene expression data. Bioinformatics 26: 139–140
Rodriguez J, Tang CHA, Khodor YL, Vodala S, Menet JS, Rosbash M (2013) Nascent-Seq
analysis of Drosophila cycling gene expression. Proc Natl Acad Sci U S A 110: E275–E284
Romanowski A, Yanovsky MJ (2015) Circadian rhythms and post-transcriptional regulation in
higher plants. Front Plant Sci 6: 437
Romero JM, Valverde F (2009) Evolutionarily conserved photoperiod mechanisms in plants:
when did plant photoperiodic signaling appear? Plant Signal Behav 4: 642–644

299

Rosa BA, Jiao YH, Oh S, Montgomery BL, Qin WS, Chen J (2012) Frequency-based timeseries gene expression recomposition using PRIISM. Bmc Syst Biol. doi: Artn 69 Doi
10.1186/1752-0509-6-69
SĂĄnchez L, Thieffry D (2001) A logical analysis of the Drosophila gap-gene system. J Theor
Biol 211: 115–141
Sanderson MJ, Thorne JL, Wikstrom N, Bremer K (2004) Molecular evidence on plant
divergence times. Am J Bot 91: 1656–1665
Santopolo S, Boccaccini A, Lorrai R, Ruta V, Capauto D, Minutello E, Serino G,
Costantino P, Vittorioso P (2015) DOF AFFECTING GERMINATION 2 is a positive
regulator of light-mediated seed germination and is repressed by DOF AFFECTING
GERMINATION 1. Plant Biol 15: 72
Schmid M, Davison TS, Henz SR, Pape UJ, Demar M, Vingron M, SchĂślkopf B, Weigel D,
Lohmann JU (2005) A gene expression map of Arabidopsis thaliana development. Nat
Genet 37: 501–506
Schnable J, Springer N, Freeling M (2011) Differentiation of the maize subgenomes by
genome dominance and both ancient and ongoing gene loss. Proc Natl Acad Sci U S A 108:
4069–4074
Schnell S (2014) Validity of the Michaelis-Menten equation--steady-state or reactant stationary
assumption: that is the question. FEBS J 281: 464–472
Schranz M, Quijada P, Sung S, Lukens L, Amasino R, Osborn T (2002) Characterization and
Effects of the Replicated Flowering Time Gene FLC in Brassica rapa. Gemetics 3: 1457–
1468
Schulze A, Downward J (2001) Navigating gene expression using microarrays--a technology
review. Nat Cell Biol 3: E190--5
Segal E, Raveh-Sadka T, Schroeder M, Unnerstall U, Gaul U (2008) Predicting expression
patterns from regulatory sequence in Drosophila segmentation. Nature 451: 535–540
Seoighe C, Gehring C (2004) Genome duplication led to highly selective expansion of the
Arabidopsis thaliana proteome. Trends Genet 20: 461–464
Shi T, Ilikchyan I, Rabouille S, Zehr JP (2010) Genome-wide analysis of diel gene expression
in the unicellular N(2)-fixing cyanobacterium Crocosphaera watsonii WH 8501. ISME J 4:
621–632
Shimada A, Dohke K, Sadaie M, Shinmyozu K, Nakayama J-I, Urano T, Murakami Y
(2009) Phosphorylation of Swi6/HP1 regulates transcriptional gene silencing at
heterochromatin. Genes Dev 23: 18–23
300

Shiu S-H, Shih M-C, Li W-H (2005) Transcription factor families have much higher expansion
rates in plants than in animals. Plant Physiol 139: 18–26
Shmulevich I, Dougherty ER, Kim S, Zhang W (2002) Probabilistic Boolean Networks: a
rule-based uncertainty model for gene regulatory networks. Bioinformatics 18: 261–274
Shmulevich I, Gluhovsky I, Hashimoto RF, Dougherty ER, Zhang W (2003) Steady-state
analysis of genetic regulatory networks modelled by probabilistic boolean networks. Comp
Funct Genomics 4: 601–608
Siaut M, Cuine S, Cagnon C, Fessler B, Nguyen M, Carrier P, Beyly A, Beisson F,
Triantaphylides C, Li-Beisson Y, et al (2011) Oil accumulation in the model green alga
Chlamydomonas reinhardtii: characterization, variability between common laboratory
strains and relationship with starch reserves. BMC Biotechnol 11: 7
Sidorova JM, Mikesell GE, Breeden LL (1995) Cell cycle-regulated phosphorylation of Swi6
controls its nuclear localization. Mol Biol Cell 6: 1641–1658
Silver D, Huang A, Maddison CJ, Guez A, Sifre L, van den Driessche G, Schrittwieser J,
Antonoglou I, Panneershelvam V, Lanctot M, et al (2016) Mastering the game of Go
with deep neural networks and tree search. Nature 529: 484–489
Singh R, Lanchantin J, Robins G, Qi Y (2016) DeepChrome: deep-learning for predicting
gene expression from histone modifications. Bioinformatics 32: i639--i648
Sinha S, Tompa M (2003) YMF: A program for discovery of novel transcription factor binding
sites by statistical overrepresentation. Nucleic Acids Res 31: 3586–3588
Soltis D, Bell C, Kim S, Soltis P (2008) Origin and early evolution of angiosperms. Ann N Y
Acad Sci 1133: 3–25
Soltis DE, Visger CJ, Soltis PS (2014) The polyploidy revolution then…and now: Stebbins
revisited. Am J Bot 101: 1057–1078
Soneson C, Delorenzi M (2013) A comparison of methods for differential expression analysis of
RNA-seq data. BMC Bioinformatics 14: 91
Song YH, Ito S, Imaizumi T (2013) Flowering time regulation: photoperiod- and temperaturesensing in leaves. Trends Plant Sci 18: 575–583
Spellman PT, Sherlock G, Zhang MQ, Iyer VR, Anders K, Eisen MB, Brown PO, Botstein
D, Futcher B (1998) Comprehensive identification of cell cycle-regulated genes of the
yeast Saccharomyces cerevisiae by microarray hybridization. Mol Biol Cell 9: 3273–3297
Spitz F, Furlong EEM (2012) Transcription factors: from enhancer binding to developmental
control. Nat Rev Genet 13: 613–626
301

Spivak AT, Stormo GD (2016) Combinatorial Cis-regulation in Saccharomyces Species. G3 6:
653–667
Stamatakis A (2006) RAxML-VI-HPC: maximum likelihood-based phylogenetic analyses with
thousands of taxa and mixed models. Bioinformatics 22: 2688–2690
Stamatakis A (2014) RAxMLversion 8: a tool for phylogenetic analysis and post-analysis of
large phylogenies. Bioinformatics 30: 1312–1313
Stormo GD, Schneider TD, Gold L, Ehrenfeucht A (1982) Use of the “Perceptron” algorithm
to distinguish translational initiation sites in E. coli. Nucleic Acids Res 10: 2997–3011
Straume M (2004) DNA microarray time series analysis: automated statistical assessment of
circadian rhythms in gene expression patterning. Methods Enzym 383: 149–166
Sukumaran S, Almon RR, DuBois DC, Jusko WJ (2010) Circadian rhythms in gene
expression: Relationship to physiology, disease, drug disposition and drug action. Adv Drug
Deliv Rev 62: 904–917
Takagi Y, Matsuda H, Taniguchi Y, Iwaisaki H (2014) Predicting the phenotypic values of
physiological traits using SNP genotype and gene expression data in mice. PLoS One 9:
e115532
Takasaki H, Maruyama K, Takahashi F, Fujita M, Yoshida T, Nakashima K, Myouga F,
Toyooka K, Yamaguchi-Shinozaki K, Shinozaki K (2015) SNAC-As, stress-responsive
NAC transcription factors, mediate ABA-inducible leaf senescence. Plant J 84: 1114–1123
Ter Linde JJM, Steensma HY (2002) A microarray-assisted screen for potential Hap1 and
Rox1 target genes in Saccharomyces cerevisiae. Yeast 19: 825–840
Teramoto H, Nakamori A, Minagawa J, Ono TA (2002) Light-intensity-dependent expression
of Lhc gene family encoding light-harvesting chlorophyll-a/b proteins of photosystem II in
Chlamydomonas reinhardtii. Plant Physiol 130: 325–333
Thomas BC, Pedersen B, Freeling M (2006) Following tetraploidy in an Arabidopsis ancestor,
genes were removed preferentially from one homeolog leaving clusters enriched in dosesensitive genes. Genome Res 16: 934–946
Thomas R (1973) Boolean formalization of genetic control circuits. J Theor Biol 42: 563–585
Tomancak P, Beaton A, Weiszmann R, Kwan E, Shu S, Lewis SE, Richards S, Ashburner
M, Hartenstein V, Celniker SE, et al (2002) Systematic determination of patterns of gene
expression during Drosophila embryogenesis. Genome Biol 3: RESEARCH0088

302

Tong AH, Evangelista M, Parsons AB, Xu H, Bader GD, PagĂŠ N, Robinson M,
Raghibizadeh S, Hogue CW, Bussey H, et al (2001) Systematic genetic analysis with
ordered arrays of yeast deletion mutants. Science 294: 2364–2368
Trabzuni D, United Kingdom Brain Expression Consortium (UKBEC), Thomson PC
(2014) Analysis of gene expression data using a linear mixed model/finite mixture model
approach: application to regional differences in the human brain. Bioinformatics 30: 1555–
1561
Trairatphisan P, Wiesinger M, Bahlawane C, Haan S, Sauter T (2016) A Probabilistic
Boolean Network Approach for the Analysis of Cancer-Specific Signalling: A Case Study
of Deregulated PDGF Signalling in GIST. PLoS One 11: e0156223
Tran L, Nakashima K, Sakuma Y, Simpson S, Fujita Y, Maruyama K, Fujita M, Seki M,
Shinozaki K, Yamaguchi-Shinozaki K (2004) Isolation and functional analysis of
Arabidopsis stress-inducible NAC transcription factors that bind to a drought-responsive
cis-element in the early responsive to dehydration stress 1 promoter. Plant Cell 16: 2481–
2498
Trapnell C, Pachter L, Salzberg SL (2009) TopHat: discovering splice junctions with RNASeq. Bioinformatics 25: 1105–1111
Trapnell C, Williams BA, Pertea G, Mortazavi A, Kwan G, van Baren MJ, Salzberg SL,
Wold BJ, Pachter L (2010) Transcript assembly and quantification by RNA-Seq reveals
unannotated transcripts and isoform switching during cell differentiation. Nat Biotechnol
28: 511–515
Truenit E, Siemering K, Hdoge S, Grbic V, Haseloff J (2006) A Map of KNAT Gene
Expression in the Arabidopsis Root. Plant Mol Biol 60: 1–20
Truernit E, Haseloff J (2007) A Role for KNAT Class II Genes in Root Development. Plant
Signal Behav 1: 10–12
Tsai ZT-Y, Lloyd JP, Shiu S-H (2017) Defining Functional Genic Regions in the Human
Genome through Integration of Biochemical, Evolutionary, and Genetic Evidence. Mol Biol
Evol 34: 1788–1798
Ueda HR, Hayashi S, Chen W, Sano M, Machida M, Shigeyoshi Y, Iino M, Hashimoto S
(2005) System-level identification of transcriptional circuits underlying mammalian
circadian clocks. Nat Genet 37: 187–192
Uygun S, Peng C, Lehti-Shiu MD, Last RL, Shiu S-H (2016) Utility and Limitations of Using
Gene Expression Data to Identify Functional Associations. PLoS Comput Biol 12:
e1005244

303

Uygun S, Seddon AE, Azodi CB, Shiu S-H (2017) Predictive Models of Spatial Transcriptional
Response to High Salinity. Plant Physiol 174: 450–464
van der Felden J, Weisser S, BrĂźckner S, Lenz P, MĂśsch H-U (2014) The transcription
factors Tec1 and Ste12 interact with coregulators Msa1 and Msa2 to activate adhesion and
multicellular development. Mol Cell Biol 34: 2283–2293
van Hoek MJA, Hogeweg P (2007) The role of mutational dynamics in genome shrinkage. Mol
Biol Evol 24: 2485–2494
van Noort V, Snel B, Huynen MA (2003) Predicting gene function by conserved co-expression.
Trends Genet 19: 238–242
van ’t Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AAM, Mao M, Peterse HL, van der
Kooy K, Marton MJ, Witteveen AT, et al (2002) Gene expression profiling predicts
clinical outcome of breast cancer. Nature 415: 530–536
Veitia RA, Bottani S, Birchler JA (2013) Gene dosage effects: nonlinearities, genetic
interactions, and dosage compensation. Trends Genet 29: 385–393
Voigt J, Munzner P (1987) The Chlamydomonas Cell-Cycle Is Regulated by a Light DarkResponsive Cell-Cycle Switch. Planta 172: 463–472
von Gromoff ED, Schroda M, Oster U, Beck CF (2006) Identification of a plastid response
element that acts as an enhancer within the Chlamydomonas HSP70A promoter. Nucleic
Acids Res 34: 4767–4779
Voss TC, Hager GL (2014) Dynamic regulation of transcriptional states by chromatin and
transcription factors. Nat Rev Genet 15: 69–81
Wang K, Wang Z, Li F, Ye W, Wang J, Song G, Yue Z, Cong L, Shang H, Zhu S, et al
(2012) The draft genome of a diploid cotton Gossypium raimondii. Nat Genet 44: 1098–
1103
Wang W, Haberer G, Gundlach H, Gläßer C, Nussbaumer T, Luo MC, Lomsadze A,
Borodovsky M, Kerstetter RA, Shanklin J, et al (2014) The Spirodela polyrhiza genome
reveals insights into its neotenous reduction fast growth and aquatic lifestyle. Nat Commun
5: 3311
Wang Z, Gerstein M, Snyder M (2009) RNA-Seq: a revolutionary tool for transcriptomics. Nat
Rev Genet 10: 57–63
Wasserman WW, Sandelin A (2004) Applied bioinformatics for the identification of regulatory
elements. Nat Rev Genet 5: 276–287

304

Weirauch M, Hughes T (2011) A catalogue of eukaryotic transcription factor types, their
evolutionary origin, and species distribution. Subcell Bichem 52: 25–73
Wichert S, Fokianos K, Strimmer K (2004) Identifying periodically expressed transcripts in
microarray time series data. Bioinformatics 20: 5–20
Winzeler EA, Shoemaker DD, Astromoff A, Liang H, Anderson K, Andre B, Bangham R,
Benito R, Boeke JD, Bussey H, et al (1999) Functional characterization of the S.
cerevisiae genome by gene deletion and parallel analysis. Science 285: 901–906
Wittenberg C, Reed SI (2005) Cell cycle-dependent transcription in yeast: promoters,
transcription factors, and transcriptomes. Oncogene 24: 2746–2755
Wolfe KH, Shields DC (1997) Molecular evidence for an ancient duplication of the entire yeast
genome. Nature 387: 708–713
Wong MKL, Krycer JR, Burchfield JG, James DE, Kuncic Z (2015) A generalised enzyme
kinetic model for predicting the behaviour of complex biochemical systems. FEBS Open
Bio 5: 226–239
Woodhouse M, Tang H, Freeling M (2011) Different Gene Families in Arabidopsis thaliana
Transposed in Different Epochs and at Different Frequencies throughout the Rosids. Plant
Cell 23: 4241–4253
Wu G, Hufnagel DE, Denton AK, Shiu S-H (2015) Retained duplicate genes in green alga
Chlamydomonas reinhardtii tend to be stress responsive and experience frequent response
gains. BMC Genomics 16: 149
Xiang CC, Chen Y (2000) cDNA microarray technology and its applications. Biotechnol Adv
18: 35–46
Yanovsky MJ, Kay SA (2002) Molecular basis of seasonal time measurement in Arabidopsis.
Nature 419: 308–312
Yeung MKS, TegnĂŠr J, Collins JJ (2002) Reverse engineering gene networks using singular
value decomposition and robust regression. Proc Natl Acad Sci U S A 99: 6163–6168
Yin L, Huang C-H, Ni J (2006) Clustering of gene expression data: performance and similarity
analysis. BMC Bioinformatics 7 Suppl 4: S19
Yuan Y, Guo L, Shen L, Liu JS (2007) Predicting gene expression from sequence: a
reexamination. PLoS Comput Biol 3: e243
Yuh CH, Bolouri H, Davidson EH (2001) Cis-regulatory logic in the endo16 gene: switching
from a specification to a differentiation mode of control. Development 128: 617–629

305

Zhang J, Tian X, Zhang H, Teng Y, Li R, Bai F, Elankumaran S, Xing J (2014) TGF-βinduced epithelial-to-mesenchymal transition proceeds through stepwise activation of
multiple feedback loops. Sci Signal 7: ra91
Zhang Y, Liu T, Meyer CA, Eeckhoute J, Johnson DS, Bernstein BE, Nusbaum C, Myers
RM, Brown M, Li W, et al (2008) Model-based analysis of ChIP-Seq (MACS). Genome
Biol 9: R137
Zhang Z, Belcram H, Magdelenat G, Couloux A, Samain S, Gill S, Rasmussen JB, Barbe V,
Faris JD, Zhang Z, et al (2012) Correction for Zhang et al., Duplication and partitioning in
evolution and function of homoeologous Q loci governing domestication characters in
polyploid wheat. Proc Natl Acad Sci 109: 1353–1353
Zheng X, Spivey N, Zeng W, Liu P, Fu Z, Klessig D, He S, Dong X (2012) Coronatine
promotes Pseudomonas syringae virulence in plants by activating a signaling cascade that
inhibits salicylic acid accumulation. Cell Host Microbe 11: 587–596
Zhu C, Byers KJRP, McCord RP, Shi Z, Berger MF, Newburger DE, Saulrieta K, Smith Z,
Shah M V, Radhakrishnan M, et al (2009) High-resolution DNA-binding specificity
analysis of yeast transcription factors. Genome Res 19: 556–566
Zhu G, Spellman PT, Volpe T, Brown PO, Botstein D, Davis TN, Futcher B (2000) Two
yeast forkhead genes regulate the cell cycle and pseudohyphal growth. Nature 406: 90–94
Zones JM, Blaby IK, Merchant SS, Umen JG (2015) High-Resolution Profiling of a
Synchronized Diurnal Transcriptome from Chlamydomonas reinhardtii Reveals Continuous
Cell and Metabolic Differentiation. Plant Cell 27: 2743–2769
Zou C, Lehti-Shiu MD, Thomashow M, Shiu SH (2009) Evolution of stress-regulated gene
expression in duplicate genes of Arabidopsis thaliana. PLoS Genet 5: e1000581
Zou C, Sun K, Mackaluso JD, Seddon AE, Jin R, Thomashow MF, Shiu SH (2011) Cisregulatory code of stress-responsive transcription in Arabidopsis thaliana. Proc Natl Acad
Sci U S A 108: 14992–14997
Zou C, Lehti-Shiu MD, Thibaud-Nissen F, Prakash T, Buell CR, Shiu S-H (2009)
Evolutionary and expression signatures of pseudogenes in Arabidopsis and rice. Plant
Physiol 151: 3–15

306