Gene regulation is thought to play a pivotal role in determiningphysiological trait variability by promoting/reducing the expressionof functional genes directly or indirectly related to the phenotype.Expression quantitative trait loci (eQTL) mapping studies hold greatpromise in disentangling gene regulations and have become a popular research area recently. In this dissertation, I explore severalstatistical strategies, which are applied to eQTL mapping studies,aimed to have a better understanding on the biological mechanism of gene regulation.The major goal of eQTL studies is to identify genomic regions thatare likely to regulate gene expressions. Given that genes functionin a network basis, we consider the scenario that a geneticperturbation could lead to a cascade effects on the transcription ofmultiple genes which belongs to a gene set, e.g., network/pathway.We develop a statistical procedure which incorporates priorbiological gene set information (e.g. KEGG pathway, GO terms) intoeQTL mapping framework to elucidate gene regulation from a systems biology perspective. Pathway regulators which mediate the expression of genes in a pathway are detected by modeling multiple gene expressions as a multivariate response to test the joint variation changes among different genotype categories. We apply the proposed approach to a yeast eQTL data set. Novel pathway regulators and regulation hotspots are identified.Currently, most eQTL mapping studies focus on single markeranalysis. However, the variability of gene expression may be causedby the regulation of a set of variants that belong to a commongenetic system, and individually only with small or moderate effect.To study the roles of genetic systems in regulating geneexpressions, we propose a statistical p-value combination approachto combine individual signals across a pre-defined genetic system to form an overall signal, while considering correlations betweengenetic variants in the system. Results for simulation studies andthe application to the yeast eQTL data are presented.As part of the DNA sequence variation, gene-gene interaction orepistasis has been ubiquitously observed in nature where its role inshaping the development of an organism has been broadly recognized. Investigating genetic interactions related to mRNA expression is an important step on the path to elucidating the genetic architecture underlying gene expression and could provide valuable functional interpretation of gene regulation. As genes are the functional units in living organisms, we conceptually propose a gene-centric gene-gene interaction framework for genome-wide epistasis detection. Multiple genetic markers (e.g. SNPs) in a gene are modeled simultaneously as a testing unit. We develop a model-based kernel machine approach for detecting pairwise gene-gene interactions. Simulation study and applications of the proposed method to the yeast eQTL data indicate its feasibility to eQTL mapping. We further extend the model-based kernel machine method to binary phenotypic outcomes. Our models provide quantitative and testable framework for assessing the interplay between gene expression and gene regulation,and will have great implications for elucidating the geneticarchitecture of gene expression.
Read
