GENOME-WIDE ASSOCIATION STUDY IN AN F2 DUROC X PIETRAIN RESOURCE POPULATION FOR MEAT QUALITY AND CARCASS TRAITS  By  Sebastián Casiró             A THESIS  Submitted to  Michigan State University in partial fulfillment of the requirements for the degree of  Animal Science-Master of Science  2016   ABSTRACT  GENOME-WIDE ASSOCIATION STUDY IN AN F2 DUROC X PIETRAIN RESOURCE POPULATION FOR MEAT QUALITY AND CARCASS TRAITS  By  Sebastián Casiró  Accurate association mapping in livestock populations is challenging, thus I tested the properties of three methods to derive 95% confidence intervals (CI) for QTL positions [Parametric Method (PM), non-parametric centered (NPC) and non-parametric non-centered (NPNC)]. The NPC failed to provide adequate coverage for the 95% CI for the true QTL position. The 95% CI obtained with NPNC and PM had similar coverage, however the PM had shorter intervals, therefore, I recommend to use PM. Furthermore, to map regions comprising SNP associated with meat quality and carcass traits I performed Genome-wide Association analysis for 948 F2 Duroc x Pietrain resource population pigs for 38 meat quality and carcass traits using 44,911 SNP. Type I error rate was controlled at a False Discovery Rate of 5%. I found nine QTL associated with 15 traits. Three of those nine QTL [one on SSC1 (tenth rib backfat thickness), one on SSC7 (dressing percentage and loin muscle area) and one on SSC11 (belly weight)] were mapped to a specific genomic segment in this study. Moreover, two novel QTL associated with tenderness were located on SSC3 and SSC5. Also, I propose the candidate genes A Kinase (PRKA) Anchor Protein 3 for the QTL on SSC5 and Carnitine O-Acetyltransferase for the QTL on SSC1. Finally, this study shows that the variants of Protein Kinase AMP-activated 3-subunit, I199V and T30N are not associated with pH 24 post mortem and related traits in this population. iii                     This thesis is gratefully dedicated to my wife Tatiana, Mom, Dad, my brothers Dami and Cheche and my grandparents Cacho, Cocho, Perla and Tamara, their love and constant support helped me to accomplish everything I proposed in my life.                        iv AKNOWLEDGMENTS   encouragement and guidance. First, I would love to show my gratitude to Drs. Alvarez, Morado, Perez Aguirreburualde, Rebuelto and Rodriguez. They always believed in me and encouraged me to pursue and advance degree. I am deeply grateful to Dr. Alvarez, who showed me the research side of veterinary medicine and has been a great counselor since I met him.  I owe my deepest gratitude to my advisor Dr. Juan Pedro Steibel, his constant guidance, encouragement and support helped me to accomplish this thesis and several projects in his lab. Furthermore, I will always treasure his advice and guidance with regards of my future goals. Also, I would love to thank my guidance committee Drs. Bates, Ernst, Lu and Steibel, who listened to my goals and suggested courses to tailor my graduate studies to achieve those goals. Moreover, I am grateful to Dr. Bates who helped me having a better understanding of the pig production in US. Also, I am grateful to Dr. Ernst who advised me throughout the candidate gene and several issues regarding the graduate student association.  I would like to thank Deborah, Kaitlin, Kaitlyn and Scott for donating their time and assisted me when I needed. Also I am grateful to Carly, Kaitlin and Kaitlyn, the  relationship that we have was crucial making this journey easier. Additionally, I want to thank Pablo and Maria, they opened the door of their house and treated me like family since the first day I arrived. Since then, they also became my extended family in East Lansing.  Also I am deeply thankful to my friends in Argentina whose daily contact made me feel I was at home. Thanks, Andy, Bari, Fede, Gri, Huevo, Juli, Leo, Lio, Pelu, Niky and Motor. Furthermore, I am grateful to my extended family in Argentina, Bety, Dani, Mica and Nati for their love and  v support since I was born. Moreover, I would love to thank Susy and Jorge, for rising such a strong, kind, funny, intelligent and lovely daughter.   I am very thankful to my entire family who always supported and encouraged me to peruse whatever I wanted to do in this life. A special thanks to my aunt Jessi who encouraged me to study abroad and advised me in every step of this journey. Thanks to Sol and Minu for the unconditional love they gave me. Also, to my grandmothers and brothers who always been there for me. Furthermore, I want to thank my parents who taught me to treasure the family and that with hard work you can accomplish every goal that you set, they are an inspiration for me. Finally, I want to thank my amazing wife Tatiana, she done the greatest sacrifice for my education. I would not be able to recover all the time that I did not spend with her these past years, and I hope I can find a way to thank her for her constant support and understanding. I think she deserves an honorary degree for everything that she done for me.   vi TABLE OF CONTENTS   LIST OF TABLES ....................................................................................................................... viii  LIST OF FIGURES ........................................................................................................................ x  CHAPTER ONE ............................................................................................................................. 1 Introduction ................................................................................................................................. 2 LITERATURE CITED ............................................................................................................... 7  CHAPTER TWO .......................................................................................................................... 12 Confidence intervals for Quantitative Trait Loci position in Genome-wide Association analysis obtained with Genomic Best Linear Unbiased Predictor models ............................... 13 ABSTRACT .............................................................................................................................. 13 Key words: ................................................................................................................................ 14 INTRODUCTION .................................................................................................................... 14 MATERIALS AND METHODS .............................................................................................. 15 Ethical Statement .................................................................................................................. 15 Real Dataset .......................................................................................................................... 15 Genotyping and genotype editing ......................................................................................... 16 Statistical Analysis ................................................................................................................ 17 Stochastic simulations ........................................................................................................... 19 Confidence interval based on a cross-validation .................................................................. 20 CI Coverage computation ..................................................................................................... 21 Post -GWAS analysis ............................................................................................................ 22 RESULTS AND DISCUSSION ............................................................................................... 22 Genome-wide Association Study in BF10 ............................................................................ 22 Confidence Interval in the real dataset.................................................................................. 23 Properties of the Confidence Intervals in the simulated dataset ........................................... 24 Efficient computational implementation of CI calculations ................................................. 28 Conclusion ............................................................................................................................ 28 LITERATURE CITED ............................................................................................................. 29  CHAPTER THREE .................................................................................................................. 33 Genome-wide association study in an F2 Duroc x Pietrain resource population for economically important meat quality and carcass traits.1 ......................................................... 34 ABSTRACT .............................................................................................................................. 35 Key words ................................................................................................................................. 36 INTRODUCTION .................................................................................................................... 36 MATERIALS AND METHODS .............................................................................................. 37 Ethical Statement .................................................................................................................. 37 Population and phenotypes ................................................................................................... 37 Genotyping and genotype editing ......................................................................................... 38  vii Genotyping of I199V and T30N ........................................................................................... 39 Statistical analysis ................................................................................................................. 39 Number of QTL per genomic region and confidence interval of peak position ................... 41 Percentage of total variance explained by the SNP .............................................................. 43 Statistical analysis for SNP in I199V and T30N .................................................................. 44 Post -GWAS analysis ............................................................................................................ 45 RESULTS ................................................................................................................................. 45 DISCUSSION ........................................................................................................................... 55 CONCLUSION ......................................................................................................................... 62 APPENDIX ............................................................................................................................... 63 LITERATURE CITED ............................................................................................................. 86  CHAPTER FOUR ......................................................................................................................... 94 Conclusion ................................................................................................................................ 95 GOALS AND CONTRIBUTIONS OF THIS STUDY ............................................................ 95 FUTURE RESEARCH DIRECTIONS: ................................................................................... 98 LITERATURE CITED ........................................................................................................... 100  viii  LIST OF TABLES   Table 2.1 Comparison between the different regions defined on chromosome 6 for tenth rib backfat thickness  Table 2.2 Summary of the confidence intervals defined on chromosome 6 for tenth rib backfat thickness in the simulated dataset   Table 3.1 Summary of the Quantitative Trait Loci regions Table 3.2 Comparison between the Genome-wide Association (GWA) peak and GWA considering I199V and T30N as fixed effects for the traits significant on SSC15   Table 3.3 Comparison of the results for the Single Nucleotide Polymorphism (SNP) peak on SSC15 and the two Protein Kinase AMP-activated 3-subunit SNP fitting equations (3.8) and (3.9)................................................................................................................................................54  Table S.1 Number of observations, phenotypic mean, phenotypic SD and heritability of the traits analyzed in this   Table S.2 Primer and reporter sequences used to genotype PRKAG3 SNP T30N and I199V.  Table S.3 List of genes in the SSC1QTL region for BF10ordered according to increasing start position expressed in bp.  Table S.4 List of genes in the SSC2 QTL region for WBS and TEN ordered according to increasing start position expressed in bp......................70  Table S.5 List of genes in the SSC5 QTL region for TEN ordered according to increasing start position expressed in bp.  Table S.6 List of genes in the SSC6 QTL region for BF10, LLBF and LW ordered according to increasing start position expressed in bp.  Table S.7 List of genes in the SSC7 QTL region associated with loin muscle area and dressing percentage, ordered according to increasing start position expressed in bp.  Table S.8 List of genes in the SSC7 QTL region associated with number of ribs and carcass length, ordered according to increasing start position expressed in bp.82   ix Table S. 9 List of genes in the SSC15 QTL region for juiciness, tenderness, WBS, 24-h pH, protein content, CY and drip loss ordered according to increasing start position expressed in bp.........................84                                        x LIST OF FIGURES   Figure 2.1 Manhattan plots for SNP associations with tenth-rib backfat thickness (BF10)..........23   Figure 2.2 Comparison of the coverage between the three methods.   Figure 3.1 Manhattan plot for SNP association with tenth-rib backfat fitting model (1) with sex, slaughter group and carcass weight as fixed effects (FE)  Figure 3.2 Manhattan plots for evaluated traits exhibiting significant QTL     1 CHAPTER ONE    2 Introduction  Traditional swine genetic improvement programs were tailored to enhance efficiency traits, such as maximizing lean growth, reducing backfat thickness and improving feed conversion to produce the highest quantity of meat at the lowest cost. Providing high value protein at a lower price is crucial for the industry; however, consumers are not only looking for quantity but also for quality, depending on their preferences and perception of the meat. Thus, the breeding goals shifted and nowadays, they are also directed towards improving meat quality. Meat quality is subject to genetic control, as revealed by between breed differences and by within breed heritabilities of relevant traits. For instance, meat quality traits heritabilities range from 0.08 to 0.30, while carcass traits heritabilities range from 0.30 to 0.60 (Sellier, 1998; van Wijk et al., 2005). Traditional breeding methods have been used to improve efficiency, carcass and meat quality traits. In particular, crossbreeding has been well exploited in swine growth and carcass traits (Schneider et al., 1982) to produce individuals with better performance than the average of their parents. Also, purebred lines have been selected to exploit within-breed variation using selection indexes (Hazel, 1943). However, selection for one trait may affect the response of correlated traits. For instance, selecting for lean growth efficiency has resulted in meat that had normal color, but was softer and more exudative (Lonergan et al., 2001). This happened because meat quality traits are negatively correlated with lean growth. For instance van Wijk et al. (2005) reported that backfat and pH 24 hours post mortem have a genetic correlation r=-0.24. Furthermore, those authors also showed that percentage of lean meat is negatively correlated with meat color traits such as Minolta L*, a* and b* in ham and loin (-0.16<r<-0.62) and with meat firmness (r= -0.21) among other traits. Likewise, van Wijk et al. (2005) estimated that   3 average daily gain is negatively correlated with meat firmness (r= -0.67). Additionally, Sellier, (1998) reported negative genetic correlations between growth traits and meat quality traits ranging from r= -0.2 to r= -0.82. Consequently, selecting for growth traits alone would lead to faster growing pigs with lower meat quality. To circumvent this problem, selection indexes should include both growth and meat quality traits. However, meat quality traits, carcass traits and some growth traits are expensive to measure because most are expressed later in life and require access to an abattoir for post-mortem data collection. This imposes a challenge to traditional breeding methods that rely on accurately estimating multi-trait breeding values from own records and from progeny records to construct selection indexes. To overcome the problem of increased cost of measuring traits later in life, having an early selection criterion would allow the swine industry to make rapid genetic progress at a reasonable cost. Currently genomic selection (Meuwissen et al., 2001; Goddard & Hayes, 2007) is widely used in livestock species (Hayes et al., 2009; Chen et al., 2011; Wang et al., 2012; Akanno et al., 2014) to estimate Genomic Breeding Values (GEBV). Genomic selection is a form of marker assisted selection, which relies on exploiting Linkage Disequilibrium (LD) between Quantitative Trait Loci (QTL) of genetic markers densely distributed along the genome (Goddard and Hayes, 2007). Thus, genomic selection enables prediction of breeding values for several traits at birth between markers and the actual causal alleles, it may have some shortcomings. For instance, the marker effects need to be re-estimated periodically (Habier et al., 2007) or marker effects estimated in one population are not informative to predict breeding values in a different population. This problem could be circumvented if the actual causative alleles, for instance, Quantitative Trait Nucleotides (QTN) are used in selection indexes (Weller and Ron, 2011).    4 Furthermore, finding QTN can facilitate the selection for negatively correlated traits, because the selection index can incorporate information about QTN that are neutral for one trait and significantly improve the other trait. For instance, several studies conducted in swine found causative variants on the Protein Kinase AMP-activated 3-subunit gene (PRKAG3) affecting meat quality (Milan et al., 2000; Ciobanu et al., 2001; Ryan et al., 2012; Uimari & Sironen, 2014), but these alleles do not seem to be associated with growth.  Finally, knowing the actual QTN behind a QTL has biotechnological value. For instance, genome editing technologies could be used to fix all desirable alleles in one generation. This technology is already being used in pigs to improve disease resistance traits (Whitworth et al., 2016) and, in combination with genomic selection, promises to significantly increase selection response (Jenko et al., 2015). Discovery of QTN of complex traits is a tedious and difficult task due to the polygenic nature of the genetic architecture of most traits (Manolio et al., 2009). A very common first step used to identify genomic regions containing QTN is to conduct a Genome-Wide Association (GWA) analysis (Weller and Ron, 2011).  In swine, the  development of  the Illumina PorcineSNP60 BeadChip (Ramos et al., 2009) allowed the implementation of genomic selection and GWA for meat quality and carcass traits (Becker et al.,2013; Ma et al., 2013; Nonneman et al., 2013; Uimari et al., 2013; Badke et al., 2014;Sanchez et al., 20 14; Bernal Rubio et al., 2015a;). A very common method used to implement GWA in livestock species in general and in swine in particular consists in the transformation of animal specific breeding values into SNP effects (Wang et al., 2012; Gualdrón Duarte et al., 2014; Bernal Rubio et al., 2015b). The output of these GWA analyses consists of a list of SNP-specific effects and/or p-values which need to be filtered for statistical significance   5 and spatially clustered to define putative QTL regions. It is expected that the QTL regions will contain the causative variants or genes behind the association signal. However accurately defining genomic regions from SNP association tests results is still challenging due to long range persistence of LD among other reasons.   In GWA different methods have been used to bound a genomic region around a QTL peak. One alternative is to select the region defined by all significant and contiguous SNP, another alternative is to consider a fixed number of SNP downstream and upstream of the peak (i.e. Hayes et al., 2010) or to use a fixed physical distance around the peak (i.e. Gualdrón Duarte et al., 2014). However, defining those fixed distances is challenging and arbitrary. A better idea is to define the region based on the LD blocks surrounding the QTL peak (Bernal Rubio et al., 2015a). But LD blocks are difficult to define objectively. Ideally, putative QTL regions should be defined using a statistical support interval. The derivation of confidence intervals (CI) has been widely studied in linkage analysis QTL mapping (Lander & Botstein, 1989; Visscher et al., 1996). For GWA analyses, Hayes (2013) proposed to calculate CI  using a data partition algorithm. However, this method has never been applied, and its properties remain unknown. Some important questions that should be answered before using the proposed CI computation methods are: Does the nominal 95% confidence intervals obtained with Hayes (2013) method provides actual 95% coverage? Is there a more efficient method that results in shorter intervals of similar coverage? These questions are important because a shorter CI will contain less genes to study as potential candidates, but at the same time, a shorter CI will have less chances of containing the actual causative gene. Consequently, implementing GWA of meat quality traits in swine using sound statistical methods that produce putative QTL regions could facilitate discovery of causative genes.   6 The overall goal of this study was to find novel QTL, to refine known QTL for meat quality and carcass traits and to propose candidate genes for further studies. To attain this overall goal, two specific aims were proposed:  1. Implement and test properties of methods for computing the confidence interval of a QTL position in the context of GWA from mixed effects GBLUP models.    2. Perform GWA of meat quality and carcass traits in an F2 Duroc x Pietrain resource population using the methods tested under aim 1 and propose candidate genes for further study.  7            LITERATURE CITED              8 LITERATURE CITED   Akanno, E. C., Plastow, G., B. W. Woodwards, S. Bauck, H. Okut, X.-L. Wu, C. Sun, J. L. Aalhus, S. S. Moore, S. P. Miller, W. Z., and J. A. Basarab. 2014. Reliability of molecular breeding values for Warner-Bratzler shear force and carcass traits of beef cattle - an independent validation study. J. Anim. Sci. 92:28962904. http://doi.org/10.2527/jas.2013-7374 Badke, Y. M., R. O. Bates, C. W. Ernst, J. Fix, and J. P. Steibel. 2014. Accuracy of estimation of genomic breeding values in pigs using low-density genotypes and imputation. G3 (Bethesda). 4:62331. http://doi.org/10.1534/g3.114.010504 Becker, D., K. Wimmers, H. Luther, A. Hofer, and T. Leeb. 2013. A Genome-Wide Association Study to Detect QTL for Commercially Important Traits in Swiss Large White Boars. PLoS One 8:16. http://doi.org/10.1371/journal.pone.0055951 Bernal Rubio, Y. L., J. L. Gualdrón Duarte, R. O. Bates, C. W. Ernst, D. Nonneman, G. A. Rohrer, A. King, S. D. Shackelford, T. L. Wheeler, R. J. C. Cantet, and J. P. Steibel. 2015a. Implementing meta-analysis from genome-wide association studies for pork quality traits 1. J. Anim. Sci. 93:56075617. http://doi.org/10.2527/jas2015-9502 Bernal Rubio, Y. L., J. L. Gualdrón Duarte, R. O. Bates, C. W. Ernst, D. Nonneman, G. A. Rohrer, A. King, S. D. Shackelford, T. L. Wheeler, R. J. C. Cantet, and J. P. Steibel. 2015b. Meta-analysis of genome-wide association from genomic prediction models. Anim. Genet.:3648. http://doi.org/10.1111/age.12378 Chen, C. Y., I. Misztal, I. Aguilar, S. Tsuruta,  the Meuwissen, S. E. Aggrey, T. Wing, and W. M. Muir. 2011. Genome-wide marker-assisted selection combining all pedigree phenotypic information with genotyping. J. Anim. Sci. 89:23. http://doi.org/10.2527/jas.2010-3071 Ciobanu, D., J. Bastiaansen, M. Malek, J. Helm, J. Woollard, G. Plastow, and M. Rothschild. 2001. Evidence for New Alleles in the Protein Kinase Adenosine Monophosphate- Activated  3 -Subunit Gene Associated With Low Glycogen Content in Pig Skeletal Muscle and Improved Meat Quality. Genetics 159:11511162. Goddard, M. E., and B. J. Hayes. 2007. Genomic selection. J.Anim.Breed.Genet 124:323330. http://doi.org/10.1080/09064700801959395 Gualdrón Duarte, J. L., R. J. C. Cantet, R. O. Bates, C. W. Ernst, N. E. Raney, and J. P. Steibel. 2014. Rapid screening for phenotype-genotype associations by linear transformations of genomic evaluations. BMC Bioinformatics 15:246. http://doi.org/10.1186/1471-2105-15-246 Habier, D., R. L. Fernando, and J. C. M. Dekkers. 2007. The impact of genetic relationship information on genome-assisted breeding values. Genetics 177:23892397. http://doi.org/10.1534/genetics.107.081190 Hayes, B. J., P. J. Bowman, A. J. Chamberlain, and M. E. Goddard. 2009. Genomic selection in dairy cattle: progress and challenges. J. Dairy Sci. 92:433443.   9 http://doi.org/10.3168/jds.2008-1646 Hayes, B. J., J. Pryce, A. J. Chamberlain, P. J. Bowman, and M. E. Goddard. 2010. Genetic architecture of complex traits and accuracy of genomic Prediction: Coat colour, Milk-fat percentage, and type in holstein cattle as contrasting model traits. PLoS Genet. 6. http://doi.org/10.1371/journal.pgen.1001139 Hayes, B. J. 2013. Overview of statistical methods for genome-wide association studies. C. Gondor, B. van der Werf, & B. J. Hayes, editors, Genome-wide association studies and genomic prediction. Human Press, New York, NY. p. 156-157. Hazel, L. N. 1943. The Genetic Basis for Constructing Selection Indexes. Genetics 28:476490. Jenko, J., G. Gorjanc, M. A. Cleveland, R. K. Varshney, C. B. A. Whitelaw, J. A. Woolliams, and J. M. Hickey. 2015. Potential of promotion of alleles by genome editing to improve quantitative traits in livestock breeding programs. Genet. Sel. Evol. 47:55. http://doi.org/10.1186/s12711-015-0135-3 Lander, E. S., and D. Botstein. 1989. Mapping Mendelian factors underlying quantitative traits using RFLP limkage maps. Genetics 121:185199. Lonergan, S. M., E. Huff-Lonergan, L. J. Rowe, D. L. Kuhlers, and S. B. Jungst. 2001. Selection for lean growth efficiency in Duroc pigs influences pork quality. J. Anim. Sci. 79:20752085. http://doi.org//2001.7982075x Ma, J., J. Yang, L. Zhou, Z. Zhang, H. Ma, X. Xie, F. Zhang, X. Xiong, L. Cui, H. Yang, X. Liu, Y. Duan, S. Xiao, H. Ai, J. Ren, and L. Huang. 2013. Genome-wide association study of meat quality traits in a White Duroc×Erhualian F2 intercross and Chinese Sutai pigs. PLoS One 8:e64047. http://doi.org/10.1371/journal.pone.0064047 Manolio, T. A., F. S. Collins, N. J. Cox, D. B. Goldstein, L. A. Hindorff, D. J. Hunter, M. I. McCarthy, E. M. Ramos, L. R. Cardon, A. Chakravarti, J. H. Cho, A. E. Guttmacher, A. Kong, L. Kruglyak, E. Mardis, C. N. Rotimi, M. Slatkin, D. Valle, A. S. Whittemore, M. Boehnke, A. G. Clark, E. E. Eichler, G. Gibson, J. L. Haines, T. F. C. Mackay, S. A. McCarroll, and P. M. Visscher. 2009. Finding the missing heritability of complex diseases. Nature 461:747753. http://doi.org/10.1038/nature08494 Meuwissen, T. H. E., B. J. Hayes, and M. E. Goddard. 2001. Prediction of total genetic value using genome-wide dense marker maps. Genetics 157:18191829. http://doi.org/11290733 Milan, D., J. T. Jeon, C. Looft, V. Amarger, A. Robic, M. Thelander, C. Rogel-Gillard, S. Paul, N. Iannuccelli, L. Rask, H. Ronne, K. Lundström, N. Reinsch, J. Gellin, E. Kalm, P. Le Roy, P. Chardon, and L. Andersson. 2000. A Mutation in PRKAG3 Associated with Excess Glycogen Content in Pig Skeletal Muscle. Science (80-. ). 288:12481251. http://doi.org/10.1126/science.288.5469.1248 Nonneman, D. J., S. D. Shackelford, D. A. King, T. L. Wheeler, R. T. Wiedmann, W. M. Snelling, and G. A. Rohrer. 2013. Genome-wide association of meat quality traits and tenderness in swine D. J. Nonneman, S. D. Shackelford, D. A. King, T. L. Wheeler, R. T.   10 Wiedmann, W. M. Snelling and G. A. Rohrer. J. Anim. Sci.:40434050. http://doi.org/10.2527/jas2013-6255 Ramos, A. M., R. P. M. A. Crooijmans, N. A. Affara, A. J. Amaral, A. L. Archibald, J. E. Beever, C. Bendixen, C. Churcher, R. Clark, P. Dehais, M. S. Hansen, J. Hedegaard, Z.-L. Hu, H. H. Kerstens, A. S. Law, H.-J. Megens, D. Milan, D. J. Nonneman, G. A. Rohrer, M. F. Rothschild, T. P. L. Smith, R. D. Schnabel, C. P. Van Tassell, J. F. Taylor, R. T. Wiedmann, L. B. Schook, and M. A. M. Groenen. 2009. Design of a High Density SNP Genotyping Assay in the Pig Using SNPs Identified and Characterized by Next Generation Sequencing Technology. PLoS One 4:e6524. http://doi.org/10.1371/journal.pone.0006524 McGee, M. Gispert, O. I. Southwood, and T. Sweeney. 2012. SNP variation in the promoter of the PRKAG3 gene and association with meat quality traits in pig. BMC Genet. 13:66. http://doi.org/10.1186/1471-2156-13-66 Sanchez, M.-P., T. Tribout, N. Iannuccelli, M. Bouffaud, B. Servin, A. Tenghe, P. Dehais, N. Muller, M. P. Del Schneider, M.-J. Mercat, C. Rogel-Gaillard, D. Milan, J.-P. Bidanel, and H. Gilbert. 2014. A genome-wide association study of production traits in a commercial population of Large White pigs: evidence of haplotypes affecting meat quality. Genet. Sel. Evol. 46:12. http://doi.org/10.1186/1297-9686-46-12 Schneider, J., L. . Christian, and D. L. Kuhlers. 1982. Crossbreeding in swine: Genetic effects on 746. Sellier, P. 1998. Genetics of meat and carcass traits. Rothschild M.F. and Ruvinsky A. editors, The genetics of the pigs. CAB International, Wallingford, UK, pp. 463-510. Uimari, P., A. Sironen, and M. L. Sevón-Aimonen. 2013. Evidence for three highly significant QTL for meat quality traits in the Finnish Yorkshire pig breed. J. Anim. Sci. 91:20012011. http://doi.org/10.2527/jas.2012-5811 Uimari, P., and A. Sironen. 2014. A combination of two variants in PRKAG3 is needed for a positive effect on meat quality in pigs. BMC Genet. 15:29. http://doi.org/10.1186/1471-2156-15-29 Visscher, P. M., R. Thompson, and C. S. Haley. 1996. Confidence Intervals in QTL Mapping by Bootstrapping. Genetics 143:10131020. Wang, H., I. Misztal, I. Aguilar, A. Legarra, and W. M. Muir. 2012. Genome-wide association mapping including phenotypes from relatives without genotypes. Genet. Res. (Camb). 94:7383. http://doi.org/10.1017/s0016672312000274 Weller, J. I., and M. Ron. 2011. Invited review: quantitative trait nucleotide determination in the era of genomic selection. J. Dairy Sci. 94:10821090. http://doi.org/68/jds.2010-3793 Whitworth, K. M., R. R. R. Rowland, C. L. Ewen, B. R. Trible, M. A. Kerrigan, A. G. Cino-Ozuna, M. S. Samuel, J. E. Lightner, D. G. McLaren, A. J. Mileham, K. D. Wells, and R. S. Prather. 2016. Gene-edited pigs are protected from porcine reproductive and respiratory   11 syndrome virus. Nat. Biotechnol. 34:2022. http://doi.org/10.1038/nbt.3434 van Wijk, H. J., D. J. G. Arts, J. O. Matthews, M. Webster, B. J. Ducro, and E. F. Knol. 2005. Genetic parameters for carcass composition and pork quality estimated in a commercial production chain. J. Anim. Sci 83:324333.   12 CHAPTER TWO    13 Confidence intervals for Quantitative Trait Loci position in Genome-wide Association analysis obtained with Genomic Best Linear Unbiased Predictor models  ABSTRACT Deriving confidence intervals (CI) for the position of quantitative trait loci has been widely studied in Linkage Analysis. However, this problem has not been fully studied for Genome-Wide Association (GWA) analyses based on mixed models. The objective of this study was to propose and test the properties of two non-parametric methods to compute CI; the non-parametric centered CI (NPC), and the non-parametric non-centered CI (NPNC). Also, we tested the properties of a previously published parametric method (PM) that, so far, had remained untested. The 10th rib backfat thickness (BF10) measurements for 947 F2 animals from the Michigan State University Duroc x Pietrain resource population were used as a base data set for simulations. We simulated 200 plasmode datasets based on the BF10 phenotypes and associated genotypes with a QTL at position 133.88Mb on SSC6 accounting for 12.6% of the genetic variation. We fitted a Gaussian linear mixed model to estimate the breeding values, and we divided the data in two halves and performed the GWA, repeating the process 200 times saving the physical position of the most significant SNP in each half-data set. Finally, we calculated the 95% CI using the empirical distribution for NPC and NPNC and assuming asymptotic normality for the PM. The 95% CI derived from the real data set showed that the NPC had the shortest interval (7.79 Mb.), followed by PM (8.84 Mb.) and NPNC (12.22 Mb.). However, the nominal 95% NPC CI only provided 89.5% coverage of the true QTL position. On the other hand, the nominal 95% PM CI and NPNC CI were slightly conservative covering 96.5% and 96% of the true QTL position. However, NPNC resulted in wider CI than the PM (9.4 vs 8 Mb.). Therefore, the PM provides   14 the shortest and most conservative CI for the QTL position. The calculation of CI with good properties (coverage and minimum length) is crucial to refine QTL regions. For instance, for BF10 the genomic region containing significantly associated SNP spanned 65.42 Mb, but it was reduced to an 8.84 Mb region using the PM CI. Having a narrower region with high probability of containing the QTL helps with candidate gene search and in the design of resequencing experiments.   Key words: optimal confidence interval, GBLUP based GWA, simulation.  INTRODUCTION Genomic selection is widely applied in livestock species to predict genomic breeding values or GEBV (Meuwissen et al., 2001; Goddard & Hayes, 2007,Hayes et al., 2009; Chen et al., 2011; Wang et al., 2012; Akanno et al., 2014). Following the estimation of the GEBV, it is common to perform Genome-Wide Association (GWA) analyses to identify SNP associated with phenotypes. A computationally efficient way of performing GWA is to estimate marker effects through linear transformation of GEBV and to estimate their variance to implement hypothesis testing (Gualdrón Duarte et al., 2014; Bernal Rubio et al., 2015). Thus, genomic regions comprising SNP associated with the traits can be mapped.  It is common that due to persistence of LD, sizable genomic regions are defined by GWA. Hence, there is a need to define a confidence interval (CI) for the true Quantitative Trait Loci (QTL) position. Refining the genomic regions by deriving the CI for a QTL narrows the list of candidate genes in a QTL region, while maximizing the chances that the actual causal genes are retained in the list. Moreover, by focusing on a narrower region, less genes can be further studied   15 to validate them as causative genes. Additionally, if further genotyping is needed, sequencing a narrower region with high confidence of containing causative genes optimizes the use of resources to characterize key genomic regions.  Therefore, accurately calculating the CI around a QTL peak is beneficial for post-GWA studies in vitro and in silico. The problem of computing the confidence interval for the position of a QTL was addressed in least squares based linkage analysis using bootstrapping (Visscher et al., 1996). However, deriving CI for the position of a QTL in a mixed model based GWA imposes a challenge: there is a need to perform randomization while accounting for the sample structure. Hayes (2013) proposed a data partition algorithm to compute the CI for the position of a QTL assuming asymptotic normality for the length of the QTL interval. However, this method has never been used and the properties of the CI using such parametric method (PM) remain unknown. The main objective of this study is to use plasmode simulations to test a previously proposed CI calculation method for GWA and to compare to non-parametric alternatives also based on randomized data-partition methods.  MATERIALS AND METHODS Ethical Statement Animal protocols were approved by the Michigan State University All University Committee on Animal Use and Care (AUF# 09/03-114-00)   Real Dataset The experimental population consisted of an F2 Duroc X Pietrain resource population created at Michigan State University (Edwards et al., 2008). Briefly, the F0 generation consisted of four   16 Duroc sires and 15 Pietrain dams. From those matings six F1 boars (representing 3 of the Duroc sires) and 50 F1 dams were selected. The F2 generation comprised a total of 1259 non inbred pigs of both sexes from 142 litters. Phenotypic data for 38 meat quality and carcass traits were recorded for approximately 948 F2 animals (Edwards et al., 2008). For illustration purposes, in this study we used backfat thickness at the 10th rib (BF10) obtained at slaughter, which is an economically important trait with a heritability of 0.449. We selected this trait because there is evidence of a QTL on SSC6 that has been reported previously by our group for this population (Edwards et al., 2008; Choi et al., 2011), and also in a Berkshire x Yorkshire population (Malek et al., 2001, Kim et al., 2005). More interestingly, the QTL region exhibits long range persistence of linkage disequilibrium, which makes it difficult to bound the QTL position. Therefore, this trait and the QTL on SSC6 is a very good example to illustrate the application of confidence interval computation methods to narrow down the position of important QTL.   Genotyping and genotype editing The animals from this experimental population were genotyped using two different SNP chips (Gualdrón Duarte et al., 2013). All the Grandparents, parents and 336 F2 animals were genotyped with the Illumina PorcineSNP60 BeadChip (Ramos et al., 2009) which encompasses 62,163 Single Nucleotide Polymorphism (SNP). The remaining 612 F2 animals were genotyped using the GeneSeek Genomic Profiler for Porcine LD (GGP-Porcine LD, GeneSeek a Neogen Company, Lincoln, NE), which is a SNP chip with lower resolution with 8,836 Tag SNP (Badke et al., 2013). We removed 2,277 SNP whose genotypes were missing for all animals in the Illumina PorcineSNP60 BeadChip. Furthermore, we checked for Mendelian inconsistencies (Forneris et al., 2015) and removed 1,155 SNP that did not fit the Mendelian inheritance rules   17 (p<8.4E-8). To account for multiple testing, the threshold (p<8.4E-8) was calculated performing a Bonferroni correction. Following these edits we imputed all the missing genotypes for both chips (Badke et al., 2013) using the FImpute software with default settings (Sargolzaei et al., 2014). During the imputation, SNP-specific accuracies where estimated having an average accuracy r2=0.97. At this point we removed 712 SNP which had an imputation accuracy r2<0.64. Moreover, 101 SNP which had more than 10% of missing genotypes were removed after the imputation, because we were unable to calculate a reliable imputation accuracy. Finally, The FImpute software detected 147 SNP and 9 animals having genotyping errors, therefore we removed those individuals and SNP from our data. Altogether the dataset used in this study has 947 F2 pigs with phenotypic records for BF10 and 44,911 SNP.  Statistical Analysis We performed a Genomic Best Linear Unbiased Predictor (GBLUP) based on a GWA, for the association study (Gualdrón Duarte et al., 2014), fitting an animal-centric Gaussian linear mixed model:  ,       (2.1) where   is the vector containing the phenotypes for BF10 in mm,  is the incidence matrix which relates the individual records with the fixed effects of sex, slaughter group and carcass weight in  (Edwards et al., 2008), ~N (0, G2A) is a vector of random breeding values where we used the marker based relationship matrix G as a covariance to account for population substructure (Janss et al., 2012). Where G was calculated by multiplying Z which is the (VanRaden, 2008) by its transpose, and Z is equal to:   18   is the allelic dosage matrix, is the allelic frequency at the jth marker, i is the ith animal and m is the number of markers. Lastly, ~N (0, I2e) is the vector of residuals. We proceeded to estimate the SNP effect () performing a linear transformation of the estimated breeding values () obtained from fitting the model in (2.1) and the estimated SNP effect variance () following (Gualdrón Duarte et al., 2014) :  ,     (2.2) ,     (2.3) where  is the portion of the inverse of the mixed model equations that correspond to animal effects,  is the inverse of the relationship matrix and the other terms have been previously described in equation (2.1). After estimating the SNP effect and its variance we calculated the test statistics by standardizing  and then p-values were obtained from the Gaussian distribution:      (2.4)  = 2(1-  (|tj|)),     (2.5) where the subscript j is the jth SNP, and  is the cumulative density function of the normal distribution. Gualdrón Duarte et al. (2014) and Bernal Rubio et al. (2015), showed that this procedure is equivalent to testing one fixed SNP at a time conditional on estimated variance ratio from model (2.1). All the computations performed in this study were done with the gwaR package (https://github.com/steibelj/gwaR.git) in the R environment (https://cran.r-project.org). To control for multiple testing a genome-wide significant threshold was determined using an FDR=5% (Benjamini & Hochberg, 1995;  Storey, 2002;  Storey & Tibshirani, 2003).    19 Stochastic simulations A stochastic simulation was performed to evaluate the properties of the confidence intervals. Genotypes from the dataset were assumed fixed (observed genotypes) and the phenotype were simulated following:   ,     (2.6)  where  is the simulated phenotype,  is the incidence matrix which relates the records with the fixed effects in . The only fixed effect used in this simulation was the general mean of the phenotypic records obtained from the real dataset, therefore  of BF10.  is a vector containing the random breeding values which are simulated from ~N (0, G 2A), where G is the genomic relationship matrix and 2A= 12.67 is the additive variance estimated from real data using model (1). ~N (0, I2e) is the simulated vector of the residuals where 2e= 15.52 is the error variance of real data estimated using model (2.1). Finally, , is the standardized genotype at marker M1GA0008917 and  is the estimated fixed effect of genotype at marker M1GA0008917 estimated from real data by fitting the following model: ,     (2.7) where  and  were previously described in equation (2.1) and  was previously explained in equation (2.6) and  is the fixed marker effect. This simulation resulted in a QTL at position 133.88 Mb, which explains 12.6% of the phenotypic variance. Furthermore, after simulating the QTL using the SNP M1GA0008917, the SNP was removed from the data, to emulate the most common situation where the causative variant itself is not genotyped. Among the remaining SNP in the data, the closest SNP (ASGA0029650) was highly correlated with the causative variant (r2=0.91)   20 Confidence interval based on a cross-validation Hayes (2013) proposed an algorithm based on data partition to compute a CI around a QTL peak in mixed model GWA. We followed his procedure, and we proposed two non-parametric alternatives. In general terms, the method consists of the following steps: Step 1: Perform a GWA with the whole dataset to estimate the QTL position: p (corresponding to the SNP with smallest p-value) Step 2: Randomly divide the dataset into two halves (, ). Step 3: Perform a GWA for each half separately, and record the physical position of the QTL peak for each half into the vectors  and  Step 4: Repeat step 2 and 3  times growing the vectors  and  in each iteration.  Calculate the CI using  and  following the proposed methods. For this study the nominal value of the CI was 95% CI.  Alternative 1: Parametric (PM) 95% Confidence Interval (Hayes, 2013) Step 5 A1: Calculate the Standard Error of the difference between the positions in v1 and v2 following:  where, k corresponds to the kth iteration from Step 4 Step 6 A1: Calculate the 95% CI using p from step 1, z value which corresponds to 97.5 percentile in a normal Gaussian distribution (z=1.96) and the ) from step 6    21  Alternative 2: Non-parametric CI centered on the maximum likelihood QTL position (NPC). Step 5 A2: Calculate .  Step 6 A2: Compute the  percentile of  (i.e. ) . Step 7 A2: Calculate the confidence interval, using p from step 1 following:   Alternative 3: Non-Parametric CI not centered on the estimated QTL position (NPNC) Step 5 A3: Reorder the pair  and  so the smallest physical position is stored in vectors  and the largest in . Where k is the kth iteration from Step 4.  Step 6 A3: Select a desired level of confidence  (i.e 95%).  Step 7 A3: Define the area under the left tail of the distribution of QTL positions outside the CI lb (e.g.: lb=0.01 for 1%).  Step 8 A3: Define the area under the right tail of the distribution of QTL positions outside the CI:  (e.g: 0.04 if lb=0.01 and c=0.95) Step 9 A3: Calculate lower boundary (LB) from lb: LB= th percentile of v1. Step 10 A3: Calculate the upper boundary (UB) from ub: UB= th percentile of v2. Step 11 A3: Calculate the length the CI ()  Step 12 A3: repeat steps 7-11 A3 for several values of lb and keep the CI of the shortest length   CI Coverage computation To calculate the coverage of the nominal 95% CI, we compare the lower and upper boundary of the  with the true QTL position, where m corresponds to the mth method and p corresponds to the pth plasmode dataset. If the true QTL position was larger than the lower boundary of the   22  and smaller than the upper boundary of the , we consider that the  covered the true QTL position. If any of those conditions were not fulfilled the  did not cover the true QTL position. Finally, the percentage of coverage was calculated as follows:  where m corresponds to the mth method, and the number of  computed was 200 in this study.  Post -GWAS analysis The genomic region used for the functionally annotated gene identification was defined by the 95% confidence interval constructed around the peak. Annotated genes within those genomic regions were identified with the ENSEMBL annotation of Sus scrofa 10.2.83 (December 2015) assembly (http://useast.ensembl.org/biomart/martview/).   RESULTS AND DISCUSSION Genome-wide Association Study in BF10 Backfat thickness is used in packing plants to determine the price of the carcass, in this study we used the BF10. This trait has a phenotypic mean (), phenotypic variance (= 53.98) and heritability () in this population. The genome-wide association study shows three QTL regions associated with the trait on SSC1, SSC6 and SSC15 (Fig. 2.1 A). For this study we concentrate on the region on SSC6 which spans from 75.14 Mb (blue arrow Fig. 2.1 B). to 140.56 Mb (red arrow Fig. 2.1 B). The genotypes of the SNP corresponding to the QTL peak (M1GA0008917 located at 133.88 Mb) explain 12.6 % of the variance. As it can be seen in Fig. 2.1 B, there appears to be at least 4 peaks on SSC6 (black arrows). Hence in order to determine how many QTL may actually explain the observed association pattern, we fitted the   23 equation (2.7), which is a Gaussian linear mixed model conditioning on the genotype of the most significant SNP on SSC6 (M1GA0008917) as a fixed effect. After fitting those genotypes as fixed effects, no other SNP in the region was associated with the trait, thus we conclude that only one QTL is present in this region. This association peak corresponds to a well known QTL previously mapped in a low resolution linkage mapping for this population (Edwards et al., 2008; Choi et al., 2011) and for other populations (Malek et al., 2001; Kim et al., 2005).  Figure 2.1 Manhattan plots for SNP associations with tenth-rib backfat thickness (BF10). Model (2.1) was fit with sex, slaughter group and carcass weight as fixed effects (FE). A, All Autosomes. B, Chromosome 6. Blue and red arrows point to the smallest and biggest physical position, respectively, of SNP associated with BF10. Black arrows point to four potential QTL peaks. Genome-wide significance threshold is shown with the blue line. False Discovery Rate (FDR<0.05)  Confidence Interval in the real dataset The results for the parametric and non-parametric approaches are shown in Table 2.1. The non-parametric centered CI has the shortest length (Table 2.1), followed by the parametric method (Table 2.1). Furthermore, the non-parametric non-centered CI is the longest (Table 2.1). Next it is important to determine if the coverage of these CI are on par with the nominal 95% level, and we do so in the next sections.   24  Table 2.1 Comparison between the different regions defined on chromosome 6 for tenth rib backfat thickness Method Lower Boundary a Upper Boundary a Length a Genes b Significant region 75.14 140.56 65.42 502 Parametric CI 129.46 138.30 8.84 42 Non-Parametric Centered CI 129.98 137.78 7.79 38 Non-Parametric Non-Centered CI  126.04 138.26 12.22 64 a Results are expressed in Mb. b Number of annotated genes in the ENSEMBLE database within those boundaries. (CI) 95% Confidence Interval  Properties of the Confidence Intervals in the simulated dataset We used simulated datasets to test the property of the CI obtained with of the three methods previously described (PM, NPC, NPNC). Figure 2.2 (A to C) shows the CI position calculated for each of the simulated datasets with each of the methods. These plots were ordered and color coded according the coverage of the CI.  Located at the top of the graphic, are those datasets where the actual QTL was covered by CI produced by the three methods, and at the bottom of the graphics is represented the dataset whose QTL were not covered by any method. CI that cover the actual QTL position are presented in black and CI that do not cover the QTL position are presented in red. None of the methods applied to dataset 1 produced a CI that covered the true position of the QTL (Fig. 2.2 A to C red arrows). For the rest of the datasets, at least one method produced CI that covered the actual QTL position.    25 The QTL region associated with BF10 in this study has more significant SNP downstream from the peak than upstream (Fig. 2.1B). This means that the persistence of LD is larger downstream from the peak. As a consequence, in any particular simulated dataset, it is more likely to have an estimated QTL peak downstream from the actual QTL position that to have the estimated QTL peak upstream from the actual QTL. Consistent with this, all the CI that did not cover the true QTL position had an upper boundary which was below the actual QTL position (133.88 Mb), following the coverage criteria explained in the materials and methods section.  The anisotropic extent of LD seems to be better captured by CI obtained with the NPNC method (Fig. 2.2 C), which tends to produce CI that extend further downstream from the estimated QTL position. The asymmetry in the CI is an advantage for the NPNC method when the estimated QTL peak position is biased further downstream. For instance, two of the simulated datasets (Fig. 2.2 A to C orange arrow) have an estimated QTL peak at position 118 and 126 Mb, clearly biased downstream from the actual QTL (133.88Mb). Interestingly, the two centered CI (PM, and NPC) calculated for those two datasets did not cover the true QTL peak position, while with CI calculated using the NPNC did cover the actual QTL position. Therefore, the NPNC CI seem to adapt better to the density of significant SNP around the peak position. However, this adaptive characteristic comes at a price in length of the interval, as we explain next. There are two properties of a CI that should be compared to determine which method is the best one; a) CI should cover the true QTL position at the nominal confidence level, b) CI should be as short as possible. The results of this comparison are shown in Table 2.2, where the average of the 200 CI calculated with the PM and NPC approaches had very similar lengths while the NPNC produced longer CI. Moreover, the 95% CI computed with PM has 96.5 % realized coverage (Table 2.2), but the 95% CI computed with the NPC had a realized coverage of only 89.5%   26 (Table 2.2). These results pose a question: how is it possible that two methods that produce CI of similar length have such a difference in the coverage? Looking in further detail at the NPC confidence intervals that do not cover the true QTL position, 16 of those had an interval length less than 1Mb centered around a peak at position 132.32Mb, while for those datasets the PM CI were on average 2.056 Mb and covered the true QTL position. On the other hand, for most datasets represented at the top of Fig. 2.2, the CI from the NPC tended to be longer than the PM CI. However, in all those datasets, the estimated QTL position was very close (<1.5Mb) to the actual QTL, so that virtually any interval at least 2 Mb in length centered at the peak contained the true position. The 95% CI obtained with NPNC had 96% realized coverage (Table 2.2), at the expense of, on average, longer intervals. Table 2.2 Summary of the confidence intervals defined on chromosome 6 for tenth rib backfat thicknes in the simulated dataset  aResults are expressed as the average values for the 200 simulated datasets in Mb. b Results are expressed in percentage.            Method Lower Boundary a Upper Boundary a Length a Coverage b Parametric 129.4 137.4 8 96.5 Non-parametric centered 129.3 137.4 8.1 89.5 Non-parametric Length Optimized 125.8 135.2 9.4 96    27   Figure 2.2 Comparison of the coverage between the three methods. The y axes represent each plasmode dataset and on the x axes is the physical position in Mb of the QTL and its confidence limits. The vertical blue line is the physical position of the real QTL (133.88 Mb.). The black line is the CI for the simulated dataset which covers the position of the real QTL. The red line is the CI for the simulated data set which does not cover the real QTL. A, CI calculated using the parametric method. B, CI calculated using the non-parametric centered method. C, CI calculated using the non-parametric non-centered method. Red arrows point to datasets whose CI do not cover the true position of the QTL with the three methods. The orange arrows point to the datasets where the estimated QTL peak was more than 10 MB apart from the true QTL peak   28 Efficient computational implementation of CI calculations  To make the CI computation faster at step 3 only a subset of markers was fitted, instead of performing a GWA with all the SNP in the dataset. This procedure is in agreement with Visscher et al., (1996), who selected markers to increase speed of the calculations. In our case, the GWA performed in each bootstrap will have half of the animals and much fewer SNP than the analysis performed in Step 1. This overall reduction in size will lower the time and memory demanded by the analysis, allowing efficient computation for several data partitions. A caveat of selecting the SNP in the QTL regions, is to add an extra checkpoint to investigate if that selection of SNP is adequate for the analysis. For instance, if the physical position saved in Step 3 are clustered at the boundaries of the selected region, a wider region is needed.   Conclusion We tested alternative methods to obtain confidence intervals for QTL positions in BLUP-based GWA. The NPC method failed to provide adequate coverage for the nominal 95% CI. The NPNC and the PM methods produced CI with almost equal coverage, but the length of CI from PM method was on average 20% shorter than the length of CI from NPNC. Neither NPNC nor PM were uniformly better, because in a few cases, the CI obtained with NPNC covered the true QTL position while the PM CI did not, and vice-versa. However, on average, both methods had similar coverage and PM produced shorter intervals. Thus, the PM CI is recommended.    29           LITERATURE CITED   30 LITERATURE CITED   Akanno, E. C., Plastow, G., B. W. Woodwards, S. Bauck, H. Okut, X.-L. Wu, C. Sun, J. L. Aalhus, S. S. Moore, S. P. Miller, W. Z., and J. A. Basarab. 2014. Reliability of molecular breeding values for Warner-Bratzler shear force and carcass traits of beef cattle - an independent validation study. J. Anim. Sci. 92:28962904. http://doi.org/10.2527/jas.2013-7374 Badke, Y. M., R. O. Bates, C. W. Ernst, C. Schwab, J. Fix, C. P. Van Tassell, and J. P. Steibel. 2013. Methods of tagSNP selection and other variables affecting imputation accuracy in swine. BMC Genet. 14:8. http://doi.org/10.1186/1471-2156-14-8 Powerful Approach to Multiple Testing When researchers tend to select pursuing multiple the ( statistically ) and support of conclusions . An unguarded use in a greatly results of single-inference inc. J.R Stat. Soc.B 57:289300. http://doi.org/10.2307/2346101 Bernal Rubio, Y. L., J. L. Gualdrón Duarte, R. O. Bates, C. W. Ernst, D. Nonneman, G. A. Rohrer, A. King, S. D. Shackelford, T. L. Wheeler, R. J. C. Cantet, and J. P. Steibel. 2015. Meta-analysis of genome-wide association from genomic prediction models. Anim. Genet.:3648. http://doi.org/10.1111/age.12378 Chen, C. Y., I. Misztal, I. Aguilar, S. Tsuruta,  the Meuwissen, S. E. Aggrey, T. Wing, and W. M. Muir. 2011. Genome-wide marker-assisted selection combining all pedigree phenotypic information with genotyping. J. Anim. Sci. 89:23. http://doi.org/10.2527/jas.2010-3071 Choi, I., J. P. Steibel, R. O. Bates, N. E. Raney, J. M. Rumph, and C. W. Ernst. 2011. Population Using Different Least-Squares Analysis Models. Front. Genet. 2:18. http://doi.org/10.3389/fgene.2011.00018 Edwards, D. B., C. W. Ernst, N. E. Raney, M. . Doumit, M. D. Hoge, and R. O. Bates. 2008. Quantitative trait loci mapping in an F2 Duroc x Pietrain resource population: II. Carcass and meat quality traits. J. Anim. Sci. 86:254:266. http://doi.org/10.2527/jas.2006-626 Forneris, N. S., A. Legarra, Z. G. Vitezica, S. Tsuruta, I. Aguilar, I. Misztal, and R. J. C. Cantet. 2015. Quality control of genotypes using heritability estimates of gene content at the marker. Genetics 199:67581. http://doi.org/10.1534/genetics.114.173559 Goddard, M. E., and B. J. Hayes. 2007. Genomic selection. J.Anim.Breed.Genet 124:323330. http://doi.org/10.1080/09064700801959395 Gualdrón Duarte, J. L., R. O. Bates, C. W. Ernst, N. E. Raney, R. J. C. Cantet, and J. P. Steibel. 2013. Genotype imputation accuracy in a F2 pig population using high density and low density SNP panels. BMC Genet. 14:38. http://doi.org/10.1186/1471-2156-14-38 Gualdrón Duarte, J. L., R. J. C. Cantet, R. O. Bates, C. W. Ernst, N. E. Raney, and J. P. Steibel.    31 2014. Rapid screening for phenotype-genotype associations by linear transformations of genomic evaluations. BMC Bioinformatics 15:246. http://doi.org/10.1186/1471-2105-15-246 Hayes, B. J., P. J. Bowman, A. J. Chamberlain, and M. E. Goddard. 2009. Genomic selection in dairy cattle: progress and challenges. J. Dairy Sci. 92:433443. http://doi.org/10.3168/jds.2008-1646 Hayes, B. J. 2013. Overview of statistical methods for genome-wide association studies. C. Gondor, B. van der Werf, & B. J. Hayes, editors, Genome-wide association studies and genomic prediction. Human Press, New York, NY. p. 156-157. Janss, L., G. de los Campos, N. Sheehan, and D. Sorensen. 2012. Inferences from genomic models in stratified populations. Genetics 192:693704. http://doi.org/10.1534/genetics.112.141143 Kim, J. J., M. F. Rothschild, J. Beever, S. Rodriguez-Zas, and J. C. M. Dekkers. 2005. Joint analysis of two breed cross populations in pigs to improve detection and characterization of quantitative trait loci. J. Anim. Sci. 83:12291240. http://doi.org/83/6/1229 Malek, M., J. C. M. Dekkers, H. K. Lee, T. J. Baas, K. Prusa, E. Huff-Lonergan, and M. F. Rothschild. 2001. A molecular genome scan analysis to identify chromosomal regions influencing economic traits in the pig. II. Meat and muscle composition. Mamm. Genome 12:637645. http://doi.org/10.1007/s003350020019 Meuwissen, T. H. E., B. J. Hayes, and M. E. Goddard. 2001. Prediction of total genetic value using genome-wide dense marker maps. Genetics 157:18191829. http://doi.org/11290733 Ramos, A. M., R. P. M. A. Crooijmans, N. A. Affara, A. J. Amaral, A. L. Archibald, J. E. Beever, C. Bendixen, C. Churcher, R. Clark, P. Dehais, M. S. Hansen, J. Hedegaard, Z.-L. Hu, H. H. Kerstens, A. S. Law, H.-J. Megens, D. Milan, D. J. Nonneman, G. A. Rohrer, M. F. Rothschild, T. P. L. Smith, R. D. Schnabel, C. P. Van Tassell, J. F. Taylor, R. T. Wiedmann, L. B. Schook, and M. A. M. Groenen. 2009. Design of a High Density SNP Genotyping Assay in the Pig Using SNPs Identified and Characterized by Next Generation Sequencing Technology. PLoS One 4:e6524. http://doi.org/10.1371/journal.pone.0006524 Sargolzaei, M., J. P. Chesnais, and F. S. Schenkel. 2014. A new approach for efficient genotype imputation using information from relatives. BMC Genomics 15:478. http://doi.org/10.1186/1471-2164-15-478 Storey, J. D., and R. Tibshirani. 2003. Statistical significance for genomewide studies. 2003. Storey, J. D. 2002. A direct approach to false discovery rates. J. R. Stat. Soc. Series B. Stat. Methodol.:479498. http://doi.org/10.1111/1467-9868.00346 VanRaden, P. M. 2008. Efficient Methods to Compute Genomic Predictions. J. Dairy Sci. 91:44144423. http://doi.org/10.3168/jds.2007-0980 Visscher, P. M., R. Thompson, and C. S. Haley. 1996. Confidence Intervals in QTL Mapping by Bootstrapping. Genetics 143:10131020.    32 Wang, H., I. Misztal, I. Aguilar, A. Legarra, and W. M. Muir. 2012. Genome-wide association mapping including phenotypes from relatives without genotypes. Genet. Res. (Camb). 94:7383. http://doi.org/10.1017/s0016672312000274    33 CHAPTER THREE     34 Genome-wide association study in an F2 Duroc x Pietrain resource population for economically important meat quality and carcass traits.1 S. Casiró*, D. Velez-Irizarry*, C.W. Ernst*, N. Raney*, R.O. Bates*, M.G. Charles* and J.P. Steibel*2 *Department of Animal Science, Michigan State University, East Lansing Department of Fisheries and Wildlife, Michigan State University. East Lansing       1This research was financially supported by the National Pork Board Grant no. 11042. Partial support was also provided by US Pig Genome Coordination funds and Fulbright Scholarship   2Corresponding author: steibelj@msu.edu   35 ABSTRACT Meat quality is essential for consumer acceptance, it ultimately impacts pork production profitability and it is subject to genetic control. The objective of this study was to map genomic regions associated with economically important meat quality and carcass traits. We performed a genome-wide association (GWA) analysis to map regions associated with 38 meat quality and carcass traits recorded for 948 F2 pigs from the Michigan State University Duroc x Pietrain resource population. The F0, F1 and 336 F2 pigs were genotyped with the Illumina Porcine SNP60 BeadChip, while the remaining F2 pigs were genotyped with the GeneSeek Genomic Profiler for Porcine LD chip, and imputed with high accuracy (r2=0.97). Altogether the genomic dataset comprised 1015 animals and 44,911 SNP. A Gaussian linear mixed model was fitted to estimate the breeding values and the variance components. A linear transformation was then performed to estimate the marker effects and variances. All the procedures were done using the gwaR package. Type I error rate was controlled at a False Discovery Rate of 5%. Seven putative QTL found in this study were previously reported in other studies. Two novel QTL associated with tenderness (TEN) were located on SSC3 (135.6:137.5Mb; FDR<0.03) and SSC5 (67.3:69.1Mb; FDR<0.02). The QTL region identified on SSC15 includes the previously reported candidate gene, Protein Kinase AMP-activated 3-subunit gene (PRKAG3), which has been associated with 24-h pH (pH24), drip loss (DL) and cook yield (CY). Also, novel candidate genes were identified for TEND in the region on SSC5 [A Kinase (PRKA) Anchor Protein 3 (AKAP3)], and for tenth rib backfat thickness (BF10) [Carnitine O-Acetyltransferase (CRAT)] in SSC1. The PRKAG3 gene has been proposed as a candidate gene for meat quality QTL on SSC15. However, there are no SNP for this gene on the chip used, thus we genotyped the animals for two non-synonymous variants (I199V and T30N). We then performed a GWA    36 conditioning on the genotype of both SNP and I199V was associated with pH24, DL, protein content (PRO) and CY (P<0.004) and T30N was associated with Juiciness, TEND, shear force, pH24, PRO and CY (P<0.04). Finally, we performed a GWA conditioning on the genotype of the SNP peak detected in this study and T30N remained associated only with PRO (P<0.02). Therefore, in this study we identified two novel QTL regions, suggest two novel candidate genes, and conclude that other SNP in PRKAG3 or a variant(s) of another nearby gene(s) explain the observed associations on SSC15 in this population.  Key words: genome-wide association, meat quality and carcass traits, PRKAG3, swine  INTRODUCTION One of the primary goals of the industry is to produce the highest quantity of meat at a lower cost, thus, they improved the lean growth. However, selecting for lean growth resulted in lower meat quality (Lonergan et al., 2001). Pork quality is judged by consumers, and whether or not they buy the product will depend on their preferences and perception of the meat. Thus, there is a need to improve meat quality in genetic improvement programs.  Genomic selection is widely used in different livestock species to predict genomic breeding values or GEBV (Meuwissen et al., 2001; Goddard & Hayes, 2007,Hayes et al., 2009; Chen et al., 2011; Wang et al., 2012; Akanno et al., 2014). GEBV prediction is usually followed by a Genome-Wide Association (GWA) analysis to identify SNP associated with traits under study. With the development of the Illumina PorcineSNP60 BeadChip (Ramos et al., 2009) several GWA studies were performed for many traits in different pig populations including reproductive traits (Becker et al., 2013), fat-related traits (Ros-Freixedes et al., 2014; Kim et al., 2015), meat    37 quality and carcass composition (Becker et al., 2013; Ma et al., 2013; Nonneman et al., 2013; Uimari et al., 2013; Sanchez et al., 2014; Bernal Rubio et al., 2015a) and growth traits (Wang et al., 2015, Gualdrón Duarte et al., 2016). Several genomic regions have been characterized for their association with economically important traits and some candidate genes have been proposed for some of these traits. A region on SSC15 identified for many populations to be associated with meat quality traits includes the gene Protein Kinase AMP-activated 3-subunit (PRKAG3). A dominant allele (R200Q) PRKAG3 was associated with glycogen content in skeletal muscle affecting the meat quality (Milan et al., 2000). Furthermore, while R200Q is only segregating in the Hampshire breed, other non-synonymous variants in PRKAG3, such as I199V and T30N were associated with improved meat quality in other breeds (Ciobanu et al., 2001). However, none of these variants are included in the Illumina PorcineSNP60 BeadChip. The goal of this study was to use the Michigan State University Duroc X Pietrain Pig Resource Population to identify new genomic regions containing Single Nucleotide Polymorphism (SNP) associated with economically important meat quality and carcass traits. Additionally, we genotyped variants I199V and T30N in PRKAG3 and we tested them for association with meat quality traits.    MATERIALS AND METHODS Ethical Statement Animal protocols were approved by the Michigan State University All University Committee on Animal Use and Care (AUF# 09/03-114-00).  Population and phenotypes The experimental population consisted of an F2 Duroc x Pietrain cross created at Michigan State    38 University and extensively described by Edwards et al. (2008). Briefly, the F0 generation consisted of four Duroc sires and 15 Pietrain dams, the F1 included 50 dams and six boars and the F2 generation comprised 954 pigs of both sexes. A total of 38 meat quality and carcass composition traits were recorded on the F2 individuals (Edwards et al., 2008). Descriptive statistics of each phenotype analyzed in this study are shown in Table S.1. The sensory panel traits tenderness (TEN) and overall tenderness (OT) were treated as a single trait (Tenderness), due to their high phenotypic correlation (r>0.97, results not shown).   Genotyping and genotype editing Two SNP chips of different densities were used to genotype the experimental population (Gualdrón Duarte et al., 2013). The entire F0, F1 and 336 F2 animals were genotyped with the Illumina PorcineSNP60 BeadChip (Ramos et al., 2009) that contains approximately 62,000 SNP. The remaining 612 F2 animals were genotyped at lower density (8,836 tag SNP) using the GeneSeek Genomic Profiler for Porcine LD (GGP-Porcine LD, GeneSeek a Neogen Company, Lincoln, NE; Badke et al., 2013). First, 2,277 SNP from the 60K chip were removed due to having genotypes missing in all animals. After that, we performed a model-based Mendelian consistency checking following Forneris et al. (2015) removing 1,155 SNP whose segregation pattern did not fit the expected Mendelian inheritance rules (p<8.4E-8). After these minimal edits we proceeded to impute all missing genotypes in the 60K chip , and those not scored in the low density chip (Badke et al., 2013). The imputation was done using the software FImpute (Sargolzaei et al., 2014), with default settings. During the imputation, SNP-specific imputation accuracies were estimated and 712 SNP were removed due to low imputation accuracy (r2<0.64). Overall imputation accuracy    39 of remaining SNP was r2=0.97. Furthermore, 101 SNP which had more than 10% of missing genotypes, were also removed after the imputation because their imputation accuracy could not be reliably estimated. The imputation algorithm flagged 147 SNP and 9 animals that contained further genotyping errors or inconsistencies. Those SNP and individuals were edited out of the genotype database. The final dataset comprised 948 F2 animals with phenotypic records and genotypes for 44,911 SNP.  Genotyping of I199V and T30N I199V and T30N are known non-synonymous substitutions from the Protein Kinase AMP-activated 3-subunit gene (PRKAG3) associated with meat quality traits such as pH 24 hours post-mortem (pH24), drip loss (DL) and cook yield (CY) ( Milan et al., 2000; Ciobanu et al., 2001). Custom Taqman genotyping assays were developed for the I199V and T30N SNP (Table S. 2). All F1 animals where genotyped and all F2 animals were either genotyped or inferred from informative homozygous F1 parents.  Statistical analysis For the association study, a GBLUP based GWAS analysis was performed (Gualdrón Duarte et al., 2014). First an animal-centric Gaussian linear mixed model was fit.       (3.1) where  is the vector containing the phenotypes,  is the incidence matrix which relates the individual records with the fixed effects of sex, slaughter group and carcass weight in  (Edwards et al., 2008), an exception was the number of ribs trait, which had sex as the only fixed effect, ~N (0, G2A) is a vector of random breeding values. The matrix G=  is the marker    40 based relationship matrix and Z matrix (VanRaden, 2008):  , where  is the allelic dosages matrix,  is the allelic frequency at the marker j of the F2 animals, i is the ith animal and m is the number of markers. The marker based relationship matrix was used to account for population substructure (Janss et al., 2012). Finally, ~N (0, I2e) is a vector of residuals; where the variance covariance I is an identity matrix. Gualdrón Duarte et al. (2014) and Bernal Rubio et al. (2015b), showed an equivalence between a test based on an animal-centric model (equation 3.1) and a test based on a SNP effects fixed model. Furthermore, we estimated the SNP effect and its variance with a linear transformation of the estimated breeding values () following Gualdrón Duarte et al. (2014):  ,     (3.2) ,    (3.3) where all the terms have been described previously in equation (3.1) except  which is the inverse of the marker based relationship matrix and  which is the portion of the inverse of the mixed model equations that correspond to animal effects. We standardized the SNP effects to obtain the test statistics:      (3.4) where the subscript j is the jth SNP. The p-values were obtained from the Gaussian distribution:  = 2(1-  (|tj|)),     (3.5)    41 were,  is the cumulative density function of the normal distribution. All computations were implemented with the gwaR package (https://github.com/steibelj/gwaR.git) in the R environment (https://cran.r-project.org). A False Discovery Rate (Benjamini & Hochberg, 1995; Storey, 2002; Storey & Tibshirani, 2003) of 5% was used as significance criteria to control for multiple tests.   Number of QTL per genomic region and confidence interval of peak position In some cases, a significant genomic region seemed to include multiple Quantitative Trait Loci (QTL) peaks. For instance, the trait tenth-rib backfat thickness (BF10) has three QTL regions on chromosomes SSC1, SSC6 and SSC15 (Fig. 3.1 A). However, the region in SSC6 shows four putative QTL peaks (red arrows in Fig. 3.1 B). In those cases, knowing the number of QTL per genomic region is necessary to calculate the 95% confidence interval of each peak. To determine the number of QTL peaks, we repeated the GWA scan but we included the genotypes of the peak SNP as a fixed effect (Fig. 3.1 C). If after fitting a SNP as a fixed effect, all other SNP in the region do not exhibit significant association, this is a strong indication of a single QTL peak in the region (Fig. 3.1 D). Additionally, if a single SNP association in another chromosome vanished when a SNP genotype in another region was included in the model (compare Fig. 3.1 A to C, green arrow), this is an indication that a single SNP is not in LD with neighboring SNP but it is in LD with many SNP on another chromosome. We did not consider such SNP for further analysis.    42  Figure 3.1 Manhattan plot for SNP association with tenth-rib backfat fitting model (1) with sex, slaughter group and carcass weight as fixed effects (FE). (A): Considering Autosomes. (B): Considering SSC6. (C): Considering autosomes and using the marker M1GA0008917 as a FE. (D): Considering SSC6 and using the marker M1GA0008917 as FE, red arrows point at four possible QTL in SSC6, green arrow shows peak in SSC15. Genome-wide significance threshold is shown with the blue line (FDR<0.05)  The 95% CI of the QTL peak position for each genomic region was computed using a method proposed by Hayes (2013). The algorithm based on cross validation comprises the following steps: Step 1:  Perform a GWA and obtain QTL peak position: p (corresponding to the SNP with the smallest p-value in a genomic region). Step 2: Randomly assign the animals in the dataset to two sets (x1, x2). Step 3: Perform a GWA analyses for x1 and x2, separately.    43 Step 4: Store the physical position of the most significant SNP in the region from x1 and x2 into the vectors v1 and v2 respectively.  Step 5: Repeat n times in order to fill v1, v2. Step 6: Calculate the Standard Error of the difference between the positions in v1 and v2 following:  where k corresponds to the position of the most significant SNP in the kth repetition.  Step 7: Calculate the 95% CI using p from step 1, z value which corresponds to 97.5 percentile in a normal Gaussian distribution (z=1.96) and the )  from step 6:   Percentage of total variance explained by the SNP The percentage of variance explained by the peak SNP was calculated by re-fitting model (3.1) including the genotypes () of the most significant SNP as a fixed effect (already described in previous section). The estimated effect of this marker was used to estimate the variance accounted for by the marker using equation (3.6).      (3.6) where  is the estimated variance associated with the marker effect,  is the genotype of the most significant marker, and b is the estimated effect of the marker. The percentage of variance explained by the marker in study can be calculated:      (3.7)    44 Where  was calculated in (3.6),  is the estimated additive genetic variance and  is the estimated error variance using the model explained in this section. The results obtained with this procedure were roughly equal to the computationally more involved methods to estimate percentage of variance explained by a QTL presented elsewhere (Hayes et al., 2010; Gualdrón Duarte et al., 2014)  Statistical analysis for SNP in I199V and T30N For the association study of the two SNP, I199V and T30N from the PRKAG3 gene, an animal-centric Gaussian linear mixed model was fitted:     (3.8) where  were previously described in model (3.1) and , are the vector of genotypes of both non-synonymous variants, expressed as the allelic dosages; counts of G (I199V) and C (T30N) alleles and  and  are the fixed effects of the markers, respectively. With this model we tested fixed SNP effects and we performed a GWA by transforming the animal effects as described in equations 3.2 to 3.5. Additionally, for the association study of the two variants conditional on the peak SNP genotype on SSC15 we fitted:    (3.9) where were previously described in model (3.1),  were previously described in model (3.8) and  is a vector containing the genotypes of the SNP with the smallest q-value on SSC15 after fitting model (3.1) and  is the marker fixed effect. Furthermore, after fitting equation 3.8, we estimated the marker effect and variance components doing a linear transformation (equations 3.2 to 3.5). The Type I Error Rate considered for this    45 analysis was  for testing the fixed effects. As previously mentioned, to account for multiple testing in the GWA scan (equation 3.8 only) a genome-wide significant threshold was determined using FDR=5%.   Post -GWAS analysis The genomic region used for the identification of candidate genes was defined by the 95% CI constructed around the peak. Annotated genes within those genomic regions were identified with the ENSEMBL annotation of Sus scrofa 10.2.83 (December 2015) assembly (http://useast.ensembl.org/biomart/martview/). We used the PigQTL database Release 28 December 2015 database (Hu et al., 2015) to approximately locate the low resolution linkage QTL detected in previous studies.  RESULTS The genome-wide association analysis found 20 putative QTL (FDR<0.05) for 15 traits. The Manhattan plots for the significant GWA analyses can be seen in Fig. 3.2.  Every region that is reported in Table 3.1 showed a single QTL peak. Some single SNP were significant (FDR<0.05), but they were not studied in more detail because they showed LD with distant QTL peaks as explained in the methods section    46    Figure 3.2 Manhattan plots for evaluated traits exhibiting significant QTL. A, Tenth rib backfat; B, WBS; C, Tenderness/OT (Overall Tenderness); D, Loin weight; E, Last lumbar vertebra backfat; F, Dressing percentage; G, Loin muscle area; H, Number of ribs; I, Carcass length; J, Belly weight; K, Protein; L, pH 24 hours post-mortem; M, Cook yield; N, Juiciness; M, Drip loss. Log10(Q-value) (y-axis) vs. SNP position (ordered within chromosome on the x-axis). The blue horizontal line marks the genome-wide significance threshold (FDR=5%)   47 Table 3.1 Summary of the Quantitative Trait Loci regions Trait Marker SSCa Posb q-valuec Effectd % Var e 95% lower pos f 95% upper pos g Genes in Region h Tenth-rib backfat ASGA0008074 1 305.0 2.5E-03 - 3.0 302.9 307.1 80 WBS M1GA0002229 2 2.9 3.3E-04 - 4.3 1.0 4.9 196  Tenderness/OT H3GA0005676 2 5.9 1.69E-04 + 4.8 4.0 7.7 Tenderness/OT H3GA0011017 3 136.5 3.21E-02 + 3.4 135.6 137.5 4 Tenderness/OT H3GA0016570 5 68.2 2.77E-02 + 3.2 67.3 69.1 16 Tenth-rib backfat M1GA0008917 6 133.9 8.65E-09 - 12.6 129.5 138.3 64 Loin weight ASGA0029651 6 133.9 1.10E-03 - 6.5 127.6 140.2 Last-lumbar vertebrae Backfat ALGA0122657 6 136.1 2.90E-03 + 5.1 131.4 140.8 Dressing Percent MARC0033464 7 35.2 1.50E-02 + 5.4 34.0 36.3 96  Loin Muscle Area ASGA0032589 7 36.4 4.60E-02 - 4.5 32.7 40.0 Carcass length ASGA0035535 7 104.0 9.80E-03 - 4.9 103.7 104.4 57  Number of ribs ALGA0043983 7 104.4 3.93E-12 - 11.7 102.5 106.2 Belly weight M1GA0015491 11 84.4 6.30E-03 - 4.5 83.6 85.2 10 Juiciness MARC0047188 15 135.2 2.60E-03 + 4.1 133.4 137.0 59  Tenderness/OT MARC0047188 15 135.2 6.71E-06 + 7.2 133.8 136.6 WBS MARC0047188 15 135.2 3.3E-04 - 5.6 134.1 136.4 24-h pH MARC0093624 15 135.5 2.36E-07 + 9.4 134.0 137.1 Drip loss MARC0093624 15 135.5 2.20E-11 - 12.8 134.9 136.1 Protein  MARC0093624 15 135.5 4.95E-20 + 21.0 135.1 135.9 Cook yield MARC0093624 15 135.5 1.55E-13 + 14.9 135.2 135.8 a Sus Scrofa Chromosome. b Peak position expressed in Megabase. c SNP q-value. d Additive effect of the SNP e Percentage of variance explained by the SNP. f Lower boundary of the 95% CI in Megabase. g Upper boundary of the 95% CI in Megabase. h Number of annotated genes in the region.   48 A SNP located at position 305 Mb on SSC1 was significantly associated with BF10. This marker explained 3% of the trait variance, with the B allele associated with less backfat thickness. The genomic region defined by the 95% CI comprised 80 genes (Table S.3). A putative candidate gene is described in the discussion section. Markers in a region on SSC2 (1.0 MB-7.7MB) were associated with two related traits: Warner-Bratzler shear force (WBS) and TEN. For WBS SNP M1GA0002229 is the most significant marker (FDR<0.001) explaining 4.3% of the phenotypic variance. The B allele was associated with lower values of the trait (B allele = less force needed to cut the chop), and the A allele was fixed in Duroc grandparents. For TEN, the QTL peak corresponded to H3GA0005676, which was 3 Mb upstream from the WBS peak, but in linkage disequilibrium (LD) with M1GA0002229 (r2=0.41) Furthermore, the 95% confidence interval for the two QTL peaks overlapped each other (Table 3.1). The B allele of the most significant marker had a positive effect on the trait (B allele = more tender chop) and the genotypes for this SNP explained 4.8% of the phenotypic variation. The whole QTL region (defined by the 95% CI) contains 196 genes (Table S.4), including three genes that were previously proposed as putative candidates: cystatin E/M (CST6; SSC2: 5.395 to 5.396 Mb), cathepsin W (CTSW; SSC2:5.550 to 5.554 Mb) and calpain-1 catalytic subunit (CAPN1; SSC2: 6.12 to 6.15). In addition to the peak on SSC2, there were other association peaks for TEN. For instance, a peak at 136.5 Mb (SNP H3GA0011017) on SSC3 (FDR<0.05) explained 3.4% of the phenotypic variance for the trait (Table 3.1). For the SNP H3GA0011017, the B allele was associated with more tender meat, and the 95% CI for the peak extended from 135.6 to 137.5 Mb (Table 3.1). This segment contained only 3 genes and one uncharacterized protein: membrane bound O-acyltransferase domain containing 2 (MBOAT2; SCC3:135.64-135.68Mb), ribonuclease L   49 (RNASEL; SSC3:135.75-135.77Mb), DNA-binding protein inhibitor ID-2 (ID2;135.801- 135.804Mb). Another novel QTL peak for TEN at position 68.2Mb on SSC5 (FDR<0.05) coincided with the SNP H3GA0016570 and it explained 3.2% of the phenotypic variance (Table 3.1). In this case the B allele was associated with increased tenderness, and it was fixed in F0 Duroc sires (f(B)=1), but was segregating in F0 Pietrain dams (f(B)=0.86). Furthermore, the 95% confidence interval region comprised 16 genes (Table S.5), including A kinase (PRKA) anchor protein 3 (AKAP3; SSC5: 68 to 68.02 Mb) located 0.2 Mb downstream the QTL peak. This gene is expressed in the longissimus dorsi muscle from the pigs in this population (data not shown). The possible relationship between this gene and TEN is further discussed in the next section. Phenotypes of three traits, BF10, last-lumbar vertebrae backfat thickness (LLBF) and loin weight (LW), were associated with SNP genotypes on SSC6. The 95% CI for those QTL peaks overlapped each other and defined a large QTL region extended between 127.6 and 140.8 Mb of SSC6 (Table 3.1). The B alleles of ASGA0029651, M1GA0008917 and ALGA0122657 were associated with a lighter LW, reduction of BF10 and increased LLBF, respectively. With regard to BF10, genotypes of SNP M1GA0008917 explained 12.6 % of the phenotypic variance. To find a SNP explaining such large proportion of variance is unusual and we performed further analyses for this SNP. The SNP M1GA0008917 was fixed in Duroc and segregating in Pietrain grandparents (f(B)=0 and f(B)=0.86 respectively) and we observed that the 25% of animals with the least backfat thickness had genotype BB with frequency 0.36 (both alleles came from the Pietrain granddames), while only 7% of the 25% of animals with the thickest backfat were BB for M1GA0008917. The Pietrain breed is well known to have less backfat than Duroc, thus these frequencies are consistent with a SNP where a common allele of Pietrain origin exerts a strong   50 effect on the phenotype. The genomic region defined for SSC6 encompasses 64 genes (Table S.6). Some candidate genes are further discussed in the next section.  Chromosome 7 had two regions associated with four traits. One region spanning from 32.7 Mb to 40 Mb contained 96 genes (Table S.7) and two markers (MARC0033464, ASGA0032589) associated (FDR<0.05) with dressing percentage (DRESS%) and loin muscle area (LMA), respectively. The B allele of the SNP MARC0033464 is associated (FDR<0.02) with higher DRESS%, while the B allele of ASGA0032589 is associated (FDR<0.05) with a smaller LMA (Table 3.1). The B allele of ASGA0032589 is fixed in Pietrain animals. Another region on SSC7 located between 102.5Mb and 106.2 Mb contained SNP associated with number of ribs (NR) and carcass length (CL). In particular, the B allele of ALGA0043983 located in this region was associated with fewer ribs.  Furthermore, genotypes of a nearby SNP ASGA003535 were associated with CL. These two markers are in LD (r2=0.6). The SNP ALGA0043983 explained 11.7% of the phenotypic variance for NR. This large proportion of explained variance was further investigated. Among the animals with 16 or more ribs, genotype AA was predominant (f(AA)= 0.64), but among animals with 13 ribs or fewer, genotype AA was the least common (f(AA)=0.10). This region contains 57 genes (Table S.8) and the QTL peak is located 1 Mb downstream of the gene vertebrae development associated gene (VRTN) that may have a substantial impact on thoracic vertebrae development, affecting the discrete trait, NR.  On SSC11, SNP located between 83.6 Mb and 85.2 Mb were associated with belly weight (FDR<0.007). The B allele of the peak SNP (M1GA0015491) was associated with lower belly weight and it explained 4.5 percent of the phenotypic variation (Table 3.1). Ten genes are located in the 95% CI for this QTL peak: myosin XVI (MYO16; SSC11:83.4-83.5Mb), collagen, type IV   51 alpha 1 and alpha 2 (COL4A1; SSC11:84.37-84.43Mb and COL4A2; SSC11:84.61:84.67), RAB20, member RAS oncogene family (RAB20; SSC11:84.68:84.7Mb) and testis expressed 29 (TEX29; SSC11:85.00:85.02Mb), and five uncharacterized proteins. Potentially relevant genes are discussed in the next section.  Chromosome 15 has a QTL region that contains markers associated with seven traits. Even though the peak SNP varied across the seven traits, the 95% confidence interval of the QTL peaks overlapped each other. Thus, we considered a single genomic region spanning from 133.4 to 137.1 Mb. The QTL peak (FDR<0.003) for juiciness (JUI), TEN and WBS corresponds to MARC0047188 (Table 3.1) where the B allele is associated with juicier and more tender meat. The marker MARC0093624 at position 135.5 Mb is associated with four traits (FDR<2.36E-07), where the B allele is associated with higher pH at 24 hours post-mortem, protein content (PRO) and CY, and with reduced drip loss (Table 3.1). The MARC0093624 SNP had breed specific allelic frequencies (Duroc: f(B)=1, Pietrain f(B)=0.86) and the percentage of variance explained by this SNP for these traits varies from 9.4% to 21% (Table 3.1). This could be explained by the genotype frequencies of phenotypically extreme animals, e.g.: the top 25% of animals with more protein were practically all of BB genotype (frequency= 95%), whereas only 38% of the bottom 25% (animals with less protein content) had BB genotypes, the most common genotype for the bottom 25% of animals with less protein content was AB (frequency = 58%). Similarly, 93% of the top 25% of animals for CY (higher cook yield) had genotype BB, while for the bottom 25% (lower cook yield), genotype BB had a frequency of 45%. Furthermore, 92% of the animals with less drip loss (bottom 25%) had a BB genotype and only 49 % of the top 25% (animals with more drip loss) were BB. Finally, this genomic region contains 59 genes (Table S.9) including   52 the Protein Kinase AMP-activated 3-subunit gene (PRKAG3; SSC15: 133.8 Mb), that we discuss in the next section, is comprised in that list.  Finally, we fit the model using equation (3.8) to determine if candidate PRKAG3 SNP were associated with the traits and if the GWA scan would still produce a genome-wide significant association peak in the region when candidate SNP genotypes are included as fixed effects in the model (Table 3.2). We found that the T30N SNP was associated with JUI, TEN, WBS, pH24, PRO and CY (P<0.01), while the I199V SNP was only associated with pH24, DL, PRO and CY (P<0.01). Furthermore, when genotypes of those candidate SNP were included as fixed effects in the GWA scan, the observed QTL peak was replicated, in some cases, in the exact position of the previous QTL peak and in other cases in a very close position with a SNP that was in high LD (r2=0.8) with the SNP in the original peak.  Table 3.2 Comparison between the Genome-wide Association (GWA) peak and GWA considering I199V and T30N as fixed effects for the traits significant on SSC15              Trait Peak GWA SNP equation (3.1) a Peak GWA SNP equation (3.8) b q-value peak GWA SNP equation (3.8) c p-value I199V d p-value T30N d Juiciness MARC0047188 MARC0047188 2.1E-02 1.30E-01 4.10E-02 Tenderness/OT MARC0047188 MARC0047188 2.6E-04 1.17E-01 1.50E-03 WBS MARC0047188 MARC0093624 4.2E-03 6.23E-02 1.00E-02 24-h pH MARC0093624 DIAS0000678 8.8E-06 4.00E-05 3.00E-02 Drip loss MARC0093624 MARC0093624 1.3E-09 6.50E-05 7.70E-02 Protein  MARC0093624 DIAS0000678 1.7E-14 2.60E-04 7.40E-09 Cook yield MARC0093624 MARC0093624 3.6E-09 4.80E-03 4.50E-03 a Name of the peak SNP when fitting the GWA without the SNP of PRKAG3 as fixed effects. b Name of the peak SNP when fitting the GWA considering I199V and T30N as fixed effects. c q-value of the peak SNP, genome-wide significance level (FDR<0.05). d p-value of PRKAG3 SNP, significant threshold (p<0.05).    53 After fitting the model using equation (3.9) and testing the significance of previously proposed SNP in PRKAG3 (Table 3.3), the only trait significantly associated with the non-synonymous variant T30N was PRO (P<0.05). The rest of the traits were not associated with either of the two non-synonymous variants of PRKAG3 evaluated (P>0.05). Moreover, including the peak SNP of the GWA explained an equal or larger proportion of the phenotypic variation than the candidate SNP (Table 3.3). For instance, without including genotype having the peak SNP from the GWA in the model (equation 3.8), the candidate SNP explained anywhere between 0.4 to 7.5% of the phenotypic variance (Table 3.3). However, once the SNP of the QTL peak was included in the model (equation 3.9), it explained from 4 to 18 % of the variation and the candidate SNP explained virtually no phenotypic variation at all (Table 3.3). In summary, we believe that the candidate SNP in PRKAG3 are not responsible for phenotypic variation for these traits in this population and that there must be other SNP in PRKAG3, or in other genes that are in LD with the peak SNP in our GWA  54 Table 3.3 Comparison of the results for the Single Nucleotide Polymorphism (SNP) peak on SSC15 and the two Protein Kinase AMP-activated 3-subunit SNP fitting equations (3.8) and (3.9)         Trait % Var. explained by peak GWA SNP equation (3.9) a  p-value I199V  equation (3.9) b % Var. explained by  I199V  equation (3.8) c % Var. explained by  I199V equation (3.9) d p-value T30N equation (3.9) b % Var. explained by T30N  equation (3.8) c  % Var. explained by T30N  equation (3.9)  d    Juiciness 4.2 9.99E-01 0.4 0.0 7.10E-01 0.7 0.0 Tenderness/OT 6.9 7.85E-01 0.5 0.0 1.59E-01 2.3 0.4 WBS 5.2 7.38E-01 0.7 0.0 4.10E-01 1.2 0.2 24-h pH 9.5 1.39E-01 3.2 0.4 5.71E-01 1.0 0.1 Drip loss 14.0 2.76E-01 3.1 0.2 1.65E-01 0.7 0.4 Protein  18.2 5.65E-01 2.5 0.1 2.00E-02 7.5 1.1 Cook yield 14.1 5.63E-01 2.4 0.1 8.44E-01 2.8 0.0 a Percentage of phenotypic variance explained by the peak SNP after performing a GWA using the model which has the SNP peak genotype and SNP PRKAG3 non-synonymous variants as a fixed effect. b p-value of the PRKAG3 SNP after fitting the model which has SNP peak and SNP PRKAG3 genotypes as fixed effects. c Percentage of phenotypic variance explained by the SNP in PRKAG3 after fitting the model which has SNP peak and SNP PRKAG3 genotypes as fixed effects. d Percentage of phenotypic variance explained by the SNP in PRKAG3 after fitting the model with the genotypes of SNP in PRKAG3 as fixed effects.    55 DISCUSSION The genomic region associated with BF10 on SSC1 (302.9-307.1 Mb), was previously reported in a Meishan x White composite population using a low resolution linkage map (Rohrer & Keelen, 1998). In this study we replicated the finding in a different population and we map it to a 3.2 Mb region using a physical position map. We found that the B allele of ASGA0008074 is associated with reduced 10th rib backfat thickness, and it is fixed in Pietrain granddams. This is consistent  with the reports that Pietrain sired animals have less backfat thickness than Duroc sired animals (Edwards et al., 2003). This region harbors the gene Carnitine O-Acetyltransferase (CRAT; SSC1:303.4-303.41 Mb), which is an enzyme that catalyzes a fully reversible exchange of acyl groups between coenzyme A and carnitine without energy consumption (Jogl et al., 2004). A previous study in beef cattle showed that the Barros ã breed had higher mRNA expression levels of CRAT than Alentejana breed in subcutaneous adipose tissue (da Costa et al., 2013). This could be associated with the storage/removal ratio of triacylglycerol (TAG) affecting fat deposition. Further studies should be carried out in order to validate CRAT as a candidate gene for this QTL.  Two SNP associated with two traits related to meat tenderness were located at the proximal end of SSC2 (1-7.7 Mb). A QTL peak for WBS (H3GA0005672, 5.90 Mb) has previously been described in a Landrace-Duroc-Yorkshire population (Nonneman et al., 2013). The SNP H3GA0005672 is located 20 Kb upstream from the peak SNP found in this study (H3GA0005676, 5.88Mb). Few studies have included sensory panel phenotypes as we have for our population, and there are no reports of QTL for TEN overlapping the region we identified on SCC2. The genomic region (1-7.7 Mb) contains previously described candidate genes including calpain-1 catalytic subunit (CAPN1) (Nonneman et al., 2013; Bernal Rubio et al., 2015a) ,   56 Cystatin E/M and Cathepsin W (Bernal Rubio et al., 2015a). The CAPN1 gene is considered as a likely candidate gene for this region (Nonneman et al., 2013). Calpain has a crucial role in post mortem changes as meat ages, degrading five key myofibrillar and cytoskeletal proteins which can contribute to post-mortem tenderization processes (Goll et al., 1992; Koohmaraie, 1992; Huff-Lonergan et al., 1996). A novel QTL region on SSC3 (135.6-137.5Mb) containing SNP associated with TEN was detected. The three genes contained in the region are: MBOAT2, RNASEL and ID2. However, there is no evidence of a biological link or apparent biological mechanism connecting these three genes and sensory panel tenderness. Hence, the genetic cause of this association peak remains unknown.  Another novel QTL region containing SNP associated with TEN was identified on SSC5 (67.3-69.1 Mb). Sixteen genes are annotated in this region including the A kinase (PRKA) anchor protein 3 (AKAP3) gene that is located 0.2 Mb downstream of the QTL peak. This gene belongs to the AKAPs (A-Kinase anchoring protein) family, which includes proteins that bind to the regulatory subunit of the adenosine monophosphate activated protein kinase (AMPK), also known as PKA (Wong and Scott, 2004). This gene has been studied mainly in sperm, and testicular/ovarian cancer. However, AKAP3 is expressed in skeletal muscle, and it is expressed in longissimus dorsi muscle in pigs in this population. The PKA enzyme plays a crucial role in glucose, glycogen and fat metabolism, and variants in the gene PRKAG3 encoding a regulatory subunit unit have been associated with meat quality traits in swine (Milan et al., 2000; Ciobanu et al., 2001; Ryan et al., 2012; Uimari & Sironen, 2014). Thus, AKAP3 is a potential candidate gene for tenderness traits. However further studies must be carried out to confirm if variants in AKAP3 cause variation in meat tenderness.    57  The QTL identified at position 127.6-140.8 Mb on SSC6 has been widely studied and well characterized for affecting not only backfat thickness, but also loin weight traits. The BF10 QTL (129.5-138.3Mb ) has been previously reported in low resolution linkage analysis in different populations ( Malek et al., 2001; Kim et al., 2005) and in this population (Edwards et al., 2008; Choi et al., 2011). In this study we confirmed and mapped it to an 8.8 Mb region of SSC6. The QTL for loin weight has previously been reported using linkage analysis (Edwards et al., 2008; Steibel et al., 2011) and here we confirmed it and mapped it to a 12.6 Mb region of SSC6 (127.6-140.2 Mb). The last lumbar vertebrae back fat thickness QTL located between 131.4 and 140.8 Mb has not been reported before. Furthermore, the genomic region defined by the three confidence intervals (127.6-140.8 Mb) includes the interval (134.6-135 Mb) described by Sanchez et al. (2014), associated with backfat traits. The genomic region defined in this study contains 64 annotated genes, where one of the most relevant genes appears to be the Leptin Receptor Overlapping Transcript (LEPROT; SSC6: 135.37-135.38), which negatively regulates leptin cell surface exposed receptors (Couturier et al., 2007). Leptin hormone has crucial roles in feed intake, growth and backfat traits. In swine, it has been shown that the serum concentrations of leptin were positively correlated with backfat thickness and negatively correlated with carcass muscle content (Berg et al., 2003). Moreover, several studies showed associations between polymorphisms in the leptin receptor (LEPR) gene and carcass measurements, including backfat thickness and loin weights (Ovilo et al., 2005; Muñoz et al., 2009; Muñoz et al., 2011;Uemoto et al., 2012). The LEPR gene which encodes multiple isoforms of the leptin receptor (Tartaglia, 1997), is located on an unassigned contig in the currently available pig genome assembly (version 10.2.83). Thus, LEPR is not included in the list of annotated genes in the QTL region (Table S.6), and it is not proposed as a candidate gene for this QTL. However, earlier studies   58 have shown that LEPR maps to this region of SSC6 (Ernst et al., 1996). Further analysis using improved genome assemblies and annotations will be needed to determine if the causative gene behind the reported QTL is LEPR, LEPROT, both of them, or another gene(s) in this region. Finally, the results of the SNP effects for M1GA0008917 (f(A)=1 in Duroc) and ALGA0122657 (f(B)=1 in Duroc) are in agreement with a previous study showing that Pietrain-sired pigs have less backfat thickness than Duroc-sired pigs (Edwards et al., 2003).  A QTL region for LMA and DRESS%, located on SSC7 (32.70-40 Mb) has already been observed this population using linkage analysis (Edwards et al., 2008; Choi et al., 2011), but results of this study refine the position of the QTL narrowing it from 27 cM and 14 cM, respectively, in the low resolution linkage maps to a specific 7.3 Mb genomic segment. The QTL for LMA has also been reported in Meishan x Pietrain and Meishan x Duroc populations using linkage analysis (Geldermann et al., 2003; Sato and Oyamada, 2003). The B allele of ASGA0032589 is fixed in Duroc grandsires and its negative substitution effect is in agreement with our previous report Edwards et al. (2003), which showed that Duroc-sired pigs had less loin muscle area than Pietrain-sired pigs. The QTL region (32.70-40 Mb) comprises 96 annotated genes, where there is no obvious biological link between those genes and the traits in this study.  The NR and CL are economically important traits. Carcass length is mainly determined by the number and length of thoracolumbar vertebras, while the number of ribs is defined by the number of thoracic vertebras. Having one more vertebrae adds on average 15 mm to the carcass length (King and Roberts, 1960), and if the extra vertebrae is thoracic it can add an extra rib, therefore the carcass will have more value than carcasses without the additional vertebra. The QTL region identified for NR and CL in this study (SSC7, 102.5-106.2 Mb) was previously reported in our linkage analysis (Edwards et al., 2008; Choi et al., 2011). According to the   59 positions reported in the PigQTL database (Hu et al., 2015) several QTL using low resolution linkage maps for carcass length (Liu et al., 2007; Uemoto & Nagamine, 2008; Yoo et al., 2014) and number of ribs ( Zhang et al., 2007) in different populations including Western, Chinese and Korean breeds were reported in this region. Also, a genome-wide association study was performed by Sanchez et al. (2014) defining a region with SNP associated with carcass length between 101.1 and 105.3 Mb, which partially overlaps with the QTL region identified in this study. Finally, of the 57 genes annotated in the QTL segment, vertebrae development associated gene (VRTN: SCC7:103.45-103.46 Mb) is a strong candidate because the number of vertebras will affect carcass length and number of ribs at the same time. Fan et al. (2013) performed a GWA study with 3 populations (White Duroc x Erhualian F2, Sutai and Erhualian x Tongcheng F2) and reported this QTL region (SSC7, 103.37-104.31 Mb) associated only with the number of the thoracic vertebrae. Additionally, these authors found 2 SNP in complete LD residing in an active promoter corresponding to two transcription binding sites in VRTN and determined that those variants were associated with the number of thoracic vertebras (Fan et al., 2013). Also, its has been shown that an insertion/deletion in the gene contributed to the carcass length and thoracic number of vertebras in a Duroc purebred population (Nakano et al., 2015). Therefore, this gene is not only affecting the number of thoracic vertebras and the carcass length, but it is also a putative candidate gene for number of ribs. Belly weight is another economically important trait, because from this primal cut, packers obtain the bacon. Thus having heavier bellies will add economic value to the carcass. We found a QTL on SCC11 (83.6-85.2 Mb) for belly weight. Our results agree with those from Milan et al. (2002), who performed a linkage analysis in a Large White x Meishan population, and reported a QTL peak in this region. We mapped this QTL region to 1.6 Mb genomic segment on SSC11.   60 The QTL region (SSC11, 83.6-85.2Mb) contains 10 genes, including COL4A1 and COL4A2 that synthetize the andchains of type IV collagen (Kühn, 1995). These two chains are the main component of the basement membranes (Van Der Rest and Garrone, 1991). Additionally, this region includes the  MYO16 which encodes Myosin XVI, involved in brain development (Patel et al., 2001). Therefore, none of the three genes in the region have been attributed functions related to muscle or fat development, or mechanisms with an obvious association with belly weight, thus the genetic cause of this association peak remains unknown.  On SSC15, QTL for seven traits were found. Because these traits are correlated, we report a single region consisting of the overlapped 95% confidence intervals (133.4-137.1 Mb). This QTL has been widely studied in different swine populations due to its relation to meat quality traits based on low resolution linkage analysis (Thomsen et al., 2004; Rohrer et al., 2005; Edwards et al., 2008; Li et al., 2010; Choi et al., 2011), GWA (Nonneman et al., 2013; Zhang et al., 2015) and in a recent meta-analysis (Bernal Rubio et al., 2015a). The QTL peak found in this study corresponding to juiciness, tenderness and WBS (135.2Mb) replicates our previous results using a low resolution map (Edwards et al., 2008; Choi et al., 2011). Additionally Thomsen et al. (2004) reported a QTL for TEN in a linkage analysis for a F2 Berkshire x Yorkshire population. A QTL peak for WBS was detected in a three way cross Duroc x Landrace x Large White, between 133-134 Mb (Zhang et al., 2015). The genome-wide study using the SNP chip allowed us to refine the QTL regions for our population, for tenderness to a 2.8 Mb region (133.8-136.6 Mb), and for juiciness to a 2.6 Mb region (133.4-137 Mb).  A QTL region associated with PRO, DL and CY has been previously reported in a linkage study for our population (Choi et al., 2011) and for CY  (Edwards et al., 2008). Also, a QTL has been reported for DL with a low resolution linkage map (Li et al., 2010), with a GWA (Zhang et al.,   61 2015) and with a meta analysis using this population  (Bernal Rubio et al., 2015a) and for CY with a low resolution map (Rohrer et al., 2005) and with an association study of SNP of PRKAG3 (Rohrer et al., 2012) and with a GWA (Nonneman et al., 2013). Although we did not refine the regions associated with WBS, PRO, DL and CY, our results add support for the region being a true QTL affecting these traits. Our report of a negative effect for DL is consistent with a pervious study showing that Duroc-sired pigs have less drip loss than Pietrain-sired pigs (Edwards et al., 2003).   The pH of the loin muscle at 24 hours post-mortem (pH24) has been widely studied, because this trait can drastically affect meat quality. The QTL for pH24 on SSC15 has been reported by several groups using low resolution linkage analysis (Edwards et al., 2008; Li et al., 2010; Choi et al., 2011), GWA (Nonneman et al., 2013; Zhang et al., 2015) and in a recent meta-analysis conducted by our group, that included the data used for this study (Bernal Rubio et al., 2015a). Some of these previous studies have proposed PRKAG3 (Choi et al., 2011; Nonneman et al., 2013; Bernal Rubio et al., 2015a; Zhang et al., 2015) as the likely candidate gene for this QTL. Moreover, specific variants within PRKAG3 have been proposed (Milan et al., 2000; Ciobanu et al., 2001; Ryan et al., 2012; Uimari and Sironen, 2014). Among the proposed candidate SNP, we genotyped I199V and T30N because according to Ciobanu et al. (2001), these two non-synonymous variants are more common to be segregating in Duroc in comparison with the G52S variant, while I199V has been associated with glycolytic potential traits in Pietrain (Ryan et al., 2012). In this study we showed that, genotypes of I199V and T30N do not fully explain the observed QTL variance. Furthermore, there are other variants in PRKAG3 not in LD with I199V and T30N that have been reported associated to pH 24 (Ryan et al., 2012; Uimari and Sironen, 2014). Therefore, other SNP in PRKAG3, or SNP in other genes in LD with MARC0093624 are   62 the causal variants behind observed phenotypic variation in pH 24 hours post-mortem and related traits in this population.   CONCLUSION We performed a GWA and used statistical support intervals to map QTL and to define genomic segments with high likelihood of containing the causative genes. We found nine QTL peaks associated with 15 traits. Two QTL associated with tenderness on SSC3 and SSC5 are novel findings. One novel candidate gene, AKAP3, was proposed for the QTL on SSC5. AKAP3 is expressed in the skeletal muscle and it binds to the regulatory subunit of PKA, thus, affecting glycogen content in the skeletal muscle, which after post mortem modification in muscle could potentially lead to inferior meat quality. The gene CRAT on SSC1, which is related to lipid metabolism was proposed as a candidate gene for BF10. Finally, we showed that the known variants I199V and T30N in the PRKAG3 gene do not fully explain the QTL found on SSC15 for pH24 and related traits.    63             APPENDIX   64 Table S.1 Number of observations, phenotypic mean, phenotypic SD and heritability of the traits analyzed in this study.           Trait n Mean SD h2  Carcass Measures      Off-farm body weight, kg 934 112.1 8.58 0.23  Hot carcass weight, kg 934 81.85 6.85 0.16  Dressing percent, % 934 73 2.12 0.24  45-min temperature, °C 933 39.42 2.16 0.07  24-h temperature, °C 931 2.9 1.19 0.15  45-min pH 920 6.37 0.22 0.09  24-h pH  913 5.51 0.14 0.18  45-min to 24-h pH decline 900 0.86 0.22 0.06  Carcass length, cm 933 78.73 2.53 0.48  Number of ribs 655 14.83 0.85 0.38  First-rib backfat, mm 845 40.62 7.06 0.23  Last-rib backfat, mm 933 28.66 6.44 0.25  Last-lumbar vertebrae backfat, mm 932 22.23 6.25 0.41  Tenth-rib backfat, mm 927 24.14 7.32 0.45  Loin Muscle Area, cm2 928 40.61 4.73 0.59 Primal cut weight      Belly weight, kg 933 5.03 0.68 0.19  Boston shoulder weight, kg 933 3.9 0.56 0.24  Ham weight, kg 933 9.63 0.77 0.5  Loin weight, kg 933 8.29 0.84 0.3  Picnic shoulder weight, kg 933 3.72 0.57 0.15  Spare rib weight , kg 930 1.53 0.2 0.38 Meat quality evaluation      a* 887 17.27 1.83 0.63  b* 887 9.1 1.61 0.2  L* 887 53.77 2.25 0.36  Color, 1 to 6 931 3.16 0.82 0.26  Firmness, 1 to 5 918 2.86 0.79 0.13  Marbling, 1 to 10 932 2.82 0.85 0.4 Proximate Analysis      Fat, % 922 3.18 1.4 0.54  Moisture, % 922 73.94 1.54 0.39  Protein, % 921 23.44 1.13 0.39   65  Laboratory analyses      Cook yield 924 77.27 2.84 0.3  Drip loss, % 932 1.83 1.17 0.27  Warner-Bratzler shear force (WBS), kg 923 3.21 0.69 0.26 Sensory panel analyses      Connective tissue, 1 to 8 928 6.38 0.39 0.1  Juiciness, 1 to 8 928 5.23 0.59 0.07  Off -flavor, 1 to 8 928 1.14 0.21 0.05  Overall tenderness (OT), 1 to 8 928 5.63 0.55 0.28  Tenderness, 1 to 8  928 5.55 0.61 0.28    66 Table S.2 Primer and reporter sequences used to genotype PRKAG3 SNP T30N and I199V.          Sequence T30N I199V Forward Primer  TGTAACCACCAGCTCAGAAAGAAG ACACCATGCTGGAGATCAAGAA Reverse Primer  CATCCTCCTGCCTTGTCCAT TGCTTCTTGCTGTCCCACAAA Reporter Sequence 1 TAGAGGCCTTGTTCCCCT CCAACGGCATCCGAG Reporter Sequence 2 AGGCCTTGGTCCCCT CCAACGGCGTCCGAG   67  Table S.3 List of genes in the SSC1QTL region for BF10ordered according to increasing start position expressed in bp.  ENSEMBLE ID Gene Name Start position  End Position  ENSSSCG00000005651 ODF2  302,880,019   302,916,932  ENSSSCG00000005650 CERCAM  302,890,305   302,898,623  ENSSSCG00000005652 GLE1  302,917,898   302,944,879  ENSSSCG00000005654 SPTAN1  302,971,843   303,025,990  ENSSSCG00000005655 WDR34  303,025,978   303,054,229  ENSSSCG00000005656 SET  303,074,195   303,088,325  ENSSSCG00000005657 PKN3  303,093,222   303,107,386  ENSSSCG00000005658 ZDHHC12  303,105,455   303,107,386  ENSSSCG00000005659 ZER1  303,110,520   303,142,117  ENSSSCG00000005660 TBC1D13  303,151,697   303,165,931  ENSSSCG00000005661 ENDOG  303,170,034   303,176,319  ENSSSCG00000005662 C9orf114  303,173,471   303,183,167  ENSSSCG00000005663 CCBL1  303,188,689   303,225,560  ENSSSCG00000005664 LRRC8A  303,225,709   303,256,403  ENSSSCG00000005665 PHYHD1  303,258,825   303,270,695  ENSSSCG00000020404 U5  303,267,226   303,267,313  ENSSSCG00000005666 DOLK  303,273,660   303,275,276  ENSSSCG00000005667 NUP188  303,275,683   303,329,083  ENSSSCG00000005668 SH3GLB2  303,329,913   303,348,472  ENSSSCG00000005669 FAM73B  303,351,066   303,380,151  ENSSSCG00000005670 DOLPP1  303,391,083   303,400,174  ENSSSCG00000005671 CRAT  303,403,630   303,417,444  ENSSSCG00000005672 PPP2R4  303,417,952   303,447,712  ENSSSCG00000005673   303,470,306   303,471,521  ENSSSCG00000005678 NTMT1  303,687,311   303,696,507  ENSSSCG00000005679 ASB6  303,696,543   303,701,842  ENSSSCG00000005680 PRRX2  303,720,888   303,770,893  ENSSSCG00000005688 PTGES  303,779,651   303,792,393  ENSSSCG00000005674 C9orf50  303,804,123   303,806,765  ENSSSCG00000005683 TOR1B  303,864,106   303,869,260  ENSSSCG00000025964 TOR1A  303,871,932   303,882,873  ENSSSCG00000030426 C9orf78  303,883,684   303,892,770  ENSSSCG00000024341 USP20  303,892,886   303,935,551    68 Table S.3  ENSSSCG00000005689 FNBP1  303,945,075   304,034,969  ENSSSCG00000005698 NCS1  304,186,077   304,192,568  ENSSSCG00000005699   304,346,787   304,445,413  ENSSSCG00000005701 ASS1  304,454,129   304,508,998  ENSSSCG00000005702 FUBP3  304,575,334   304,625,560  ENSSSCG00000005703 PRDM12  304,646,521   304,662,114  ENSSSCG00000005704 EXOSC2  304,671,176   304,679,921  ENSSSCG00000005706 ABL1  304,685,889   304,829,672  ENSSSCG00000005705 QRFP  304,834,824   304,835,228  ENSSSCG00000005707 FIBCD1  304,844,674   304,879,017  ENSSSCG00000005710 LAMC3  304,979,408   305,098,340  ENSSSCG00000005711 NUP214  305,131,025   305,241,772  ENSSSCG00000028172 FAM78A  305,269,556   305,284,180  ENSSSCG00000005713 PLPP7  305,296,657   305,297,166  ENSSSCG00000022130   305,352,962   305,353,855  ENSSSCG00000005715   305,377,245   305,439,325  ENSSSCG00000029718 SNORD62  305,426,341   305,426,426  ENSSSCG00000021682 SNORD62  305,430,346   305,430,431  ENSSSCG00000005716 POMT1  305,440,958   305,456,203  ENSSSCG00000005717 UCK1  305,455,972   305,462,254  ENSSSCG00000027241   305,471,374   305,486,208  ENSSSCG00000005719 RAPGEF1  305,502,184   305,633,144  ENSSSCG00000005720 MED27  305,733,695   305,937,230  ENSSSCG00000005721   306,021,149   306,021,451  ENSSSCG00000005723 NTNG2  306,096,986   306,137,454  ENSSSCG00000005724 SETX  306,154,772   306,228,908  ENSSSCG00000005725 TTF1  306,231,645   306,244,293  ENSSSCG00000026458   306,284,353   306,285,021  ENSSSCG00000005727 CFAP77  306,333,931   306,368,383  ENSSSCG00000005730 BARHL1  306,405,795   306,412,475  ENSSSCG00000005729 DDX31  306,414,455   306,486,475  ENSSSCG00000005731 GTF3C4  306,486,802   306,506,690  ENSSSCG00000005732 AK8  306,530,099   306,530,326  ENSSSCG00000023474   306,637,564   306,641,266  ENSSSCG00000027595   306,696,779   306,698,101  ENSSSCG00000005738 RALGDS  306,700,486   306,714,090  ENSSSCG00000028671 CEL  306,719,763   306,726,596  ENSSSCG00000005739 GTF3C5  306,729,965   306,927,497    69 Table S.3  ENSSSCG00000022250   306,779,660   306,788,316  ENSSSCG00000005733   306,794,635   306,820,070  ENSSSCG00000005737 GFI1B  306,853,465   306,857,748  ENSSSCG00000005735   306,897,871   306,903,460  ENSSSCG00000005734   306,919,984   306,923,953  ENSSSCG00000023173 REXO4  306,985,377   306,995,916  ENSSSCG00000021241 ADAMTS13  306,998,075   307,029,937  ENSSSCG00000021113 CACFD1  307,030,425   307,039,181  ENSSSCG00000024166 SLC2A6  307,041,086   307,049,206    70  Table S.4 List of genes in the SSC2 QTL region for WBS and TEN ordered according to increasing start position expressed in bp.  ENSEMBLE ID Gene Name Start position  End Position  ENSSSCG00000022736   1,051,307   1,071,874  ENSSSCG00000012864   1,228,857   1,235,210  ENSSSCG00000030888 FADD  1,388,619   1,390,090  ENSSSCG00000022579   1,469,314   1,477,594  ENSSSCG00000022123   1,487,597   1,488,346  ENSSSCG00000021469   1,508,154   1,518,080  ENSSSCG00000029328   1,563,459   1,567,755  ENSSSCG00000024158   1,575,575   1,605,841  ENSSSCG00000029868 ANO1  1,621,019   1,651,965  ENSSSCG00000012869   1,866,961   1,902,581  ENSSSCG00000012872 FGF3  1,965,237   1,973,613  ENSSSCG00000012870 FGF4  2,004,484   2,006,245  ENSSSCG00000012871 FGF19  2,060,948   2,064,674  ENSSSCG00000012874 ORAOV1  2,095,043   2,103,105  ENSSSCG00000012873 CCND1  2,342,270   2,343,202  ENSSSCG00000012875 TPCN2  2,708,675   2,727,191  ENSSSCG00000012878 IGHMBP2  2,751,383   2,774,853  ENSSSCG00000012879 MRPL21  2,773,574   2,786,091  ENSSSCG00000012880 CPT1A  2,815,220   2,874,163  ENSSSCG00000012881   2,828,172   2,838,805  ENSSSCG00000012882 MTL5  2,878,052   2,920,326  ENSSSCG00000012883 GAL  2,931,347   2,937,425  ENSSSCG00000012884 PPP6R3  2,974,853   3,097,590  ENSSSCG00000012885 LRP5  3,168,524   3,213,765  ENSSSCG00000012886   3,247,137   3,257,948  ENSSSCG00000012887   3,294,895   3,332,625  ENSSSCG00000012889 CHKA  3,416,545   3,473,335  ENSSSCG00000012890 TCIRG1  3,476,378   3,488,420  ENSSSCG00000023420   3,498,054   3,507,821  ENSSSCG00000028501   3,518,994   3,536,523  ENSSSCG00000012893   3,539,587   3,551,073  ENSSSCG00000026349   3,563,353   3,568,662  ENSSSCG00000012895   3,594,982   3,598,550    71 Table S.4  ENSSSCG00000012896 NDUFV1  3,598,806   3,604,500  ENSSSCG00000012897 GSTP1  3,618,668   3,620,934  ENSSSCG00000012900 CABP2  3,621,643   3,624,689  ENSSSCG00000012901   3,634,981   3,637,655  ENSSSCG00000028107   3,724,619   3,727,474  ENSSSCG00000020764   3,734,833   3,739,060  ENSSSCG00000012902 CDK2AP2  3,744,053   3,746,289  ENSSSCG00000012903 PITPNM1  3,748,579   3,760,628  ENSSSCG00000012904 AIP  3,761,548   3,767,520  ENSSSCG00000012905 TMEM134  3,775,765   3,782,603  ENSSSCG00000012906 CABP4  3,787,774   3,790,770  ENSSSCG00000012920 CORO1B  3,799,100   3,804,062  ENSSSCG00000012909 PTPRCAP  3,806,257   3,806,883  ENSSSCG00000012910 RPS6KB2  3,807,219   3,813,390  ENSSSCG00000012911 CARNS1  3,814,979   3,821,634  ENSSSCG00000012912 TBC1D10C  3,830,957   3,837,376  ENSSSCG00000012914 RAD9A  3,838,643   3,848,285  ENSSSCG00000012913 PPP1CA  3,838,643   3,843,037  ENSSSCG00000012915 CLCF1  3,885,576   3,888,922  ENSSSCG00000012916   3,944,557   3,947,680  ENSSSCG00000027956 SSH3  3,968,534   3,976,217  ENSSSCG00000012917 ANKRD13D  3,976,503   3,990,125  ENSSSCG00000012918 ADRBK1  3,992,424   4,000,951  ENSSSCG00000029637   4,021,430   4,051,218  ENSSSCG00000021760   4,156,989   4,182,781  ENSSSCG00000022726 RHOD  4,233,085   4,238,597  ENSSSCG00000025443   4,350,835   4,351,810  ENSSSCG00000024837 SYT12  4,385,769   4,409,341  ENSSSCG00000028474   4,514,069   4,515,044  ENSSSCG00000012925   4,576,317   4,585,943  ENSSSCG00000012926   4,604,883   4,609,731  ENSSSCG00000012927 RCE1  4,612,347   4,615,044  ENSSSCG00000029902 C11orf80  4,615,256   4,674,876  ENSSSCG00000029317   4,752,997   4,792,821  ENSSSCG00000012946   4,793,528   4,810,745  ENSSSCG00000021891   4,816,099   4,825,400  ENSSSCG00000025136   4,826,321   4,839,557  ENSSSCG00000021340   4,839,809   4,844,997    72 Table S.4  ENSSSCG00000012933 CCDC87  4,864,531   4,867,104  ENSSSCG00000012934   4,867,366   4,875,351  ENSSSCG00000027368   4,930,949   4,936,147  ENSSSCG00000030087   4,936,399   4,949,636  ENSSSCG00000012947   4,950,491   4,957,808  ENSSSCG00000025339   4,965,113   4,982,609  ENSSSCG00000030415   4,983,355   5,022,966  ENSSSCG00000012944 PELI3  5,026,250   5,036,536  ENSSSCG00000012943 MRPL11  5,054,876   5,057,918  ENSSSCG00000012942 NPAS4  5,067,608   5,072,359  ENSSSCG00000012941 SLC29A2  5,104,241   5,111,869  ENSSSCG00000012940 B4GAT1  5,120,057   5,122,388  ENSSSCG00000012939 BRMS1  5,122,632   5,130,266  ENSSSCG00000012950 RIN1  5,131,242   5,135,843  ENSSSCG00000029949 CD248  5,146,889   5,149,543  ENSSSCG00000012952 TMEM151A  5,166,663   5,171,830  ENSSSCG00000012953 YIF1A  5,173,983   5,178,254  ENSSSCG00000012954 CNIH2  5,178,535   5,184,499  ENSSSCG00000012955 KLC2  5,189,928   5,198,860  ENSSSCG00000012956 PACS1  5,210,217   5,246,316  ENSSSCG00000012957 SF3B2  5,351,240   5,366,039  ENSSSCG00000028015 GAL3ST3  5,373,567   5,378,222  ENSSSCG00000012959 CATSPER1  5,384,927   5,393,145  ENSSSCG00000012960 CST6  5,395,320   5,396,971  ENSSSCG00000012961   5,405,034   5,405,903  ENSSSCG00000027334   5,408,048   5,411,734  ENSSSCG00000029016   5,420,795   5,425,387  ENSSSCG00000027377   5,437,528   5,438,439  ENSSSCG00000012962   5,440,257   5,441,959  ENSSSCG00000012963 SART1  5,458,386   5,477,036  ENSSSCG00000012964 TSGA10IP  5,478,348   5,491,342  ENSSSCG00000012965 DRAP1  5,513,474   5,515,508  ENSSSCG00000012966 C11orf68  5,515,886   5,518,146  ENSSSCG00000012967 FOSL1  5,534,025   5,541,866  ENSSSCG00000012968 CCDC85B  5,541,055   5,541,663  ENSSSCG00000012969 FIBP  5,544,493   5,553,916  ENSSSCG00000012970 CTSW  5,550,132   5,554,197  ENSSSCG00000012971 EFEMP2  5,558,977   5,570,871    73 Table S.4  ENSSSCG00000012973 MUS81  5,565,606   5,572,868  ENSSSCG00000012974 CFL1  5,572,133   5,579,066  ENSSSCG00000012975 SNX32  5,580,145   5,584,189  ENSSSCG00000027361 OVOL1  5,630,671   5,642,984  ENSSSCG00000012977   5,646,350   5,646,901  ENSSSCG00000012978 AP5B1  5,652,042   5,655,474  ENSSSCG00000012980 KAT5  5,666,205   5,676,732  ENSSSCG00000012979 RNASEH2C  5,667,175   5,668,321  ENSSSCG00000012981 RELA  5,698,668   5,707,943  ENSSSCG00000012982   5,760,306   5,768,449  ENSSSCG00000012983 PCNXL3  5,771,503   5,791,589  ENSSSCG00000012984 SCYL1  5,830,476   5,846,197  ENSSSCG00000012985 LTBP3  5,840,778   5,861,803  ENSSSCG00000012988 EHBP1L1  5,865,073   5,880,838  ENSSSCG00000012986   5,881,656   5,884,689  ENSSSCG00000020737   5,904,731   5,907,469  ENSSSCG00000012992 FRMD8  5,955,700   5,976,369  ENSSSCG00000012993 SLC25A45  5,984,015   5,990,284  ENSSSCG00000012994 TIGD3  5,999,046   6,000,491  ENSSSCG00000012995 DPF2  6,003,983   6,020,495  ENSSSCG00000012996 CDC42EP2  6,030,298   6,036,364  ENSSSCG00000012997 POLA2  6,054,435   6,081,792  ENSSSCG00000012998   6,102,436   6,127,700  ENSSSCG00000012999 CAPN1  6,129,549   6,155,373  ENSSSCG00000030977 CU457406.2  6,143,457   6,143,931  ENSSSCG00000019315 U6  6,182,339   6,182,442  ENSSSCG00000027057 SYVN1  6,192,447   6,200,950  ENSSSCG00000013001 MRPL49  6,198,564   6,203,040  ENSSSCG00000013002   6,203,200   6,204,847  ENSSSCG00000013003 ZNHIT2  6,207,436   6,208,686  ENSSSCG00000013004 TM7SF2  6,208,815   6,213,504  ENSSSCG00000013005 VPS51  6,213,604   6,226,834  ENSSSCG00000013006 TMEM262  6,230,658   6,231,296  ENSSSCG00000013007 ZFPL1  6,231,271   6,235,566  ENSSSCG00000013008 CDCA5  6,235,700   6,251,633  ENSSSCG00000013010 NAALADL1  6,259,685   6,269,977  ENSSSCG00000013011 SAC3D1  6,270,450   6,273,674  ENSSSCG00000013012 SNX15  6,275,115   6,287,159    74 Table S.4  ENSSSCG00000013013 ARL2  6,290,009   6,299,042  ENSSSCG00000023178 BATF2  6,312,310   6,321,081  ENSSSCG00000013015 GPHA2  6,378,546   6,379,201  ENSSSCG00000013016 PPP2R5B  6,379,401   6,386,952  ENSSSCG00000023890 ATG2A  6,393,343   6,413,661  ENSSSCG00000019906 ssc-mir-194a  6,416,776   6,416,859  ENSSSCG00000018349 ssc-mir-192  6,416,980   6,417,059  ENSSSCG00000013017 EHD1  6,430,537   6,452,599  ENSSSCG00000013018 CDC42BPG  6,460,653   6,480,139  ENSSSCG00000013019 MEN1  6,490,721   6,497,546  ENSSSCG00000024968   6,497,782   6,497,852  ENSSSCG00000013020 MAP4K2  6,497,805   6,512,966  ENSSSCG00000013021 SF1  6,524,030   6,536,326  ENSSSCG00000013022 PYGM  6,542,313   6,553,842  ENSSSCG00000013023 RASGRP2  6,557,414   6,573,068  ENSSSCG00000013024 NRXN2  6,587,432   6,684,684  ENSSSCG00000013025 SLC22A12  6,688,308   6,696,328  ENSSSCG00000020831 SLC22A11  6,777,940   6,791,460  ENSSSCG00000013030 PRDX5  6,864,027   6,867,851  ENSSSCG00000013028 ESRRA  6,867,951   6,878,759  ENSSSCG00000013029 TRMT112  6,868,094   6,869,375  ENSSSCG00000013027 TEX40  6,881,439   6,883,031  ENSSSCG00000013031 KCNK4  6,883,965   6,890,656  ENSSSCG00000013032 GPR137  6,893,208   6,896,116  ENSSSCG00000013033 BAD  6,897,945   6,909,698  ENSSSCG00000013034 PLCB3  6,911,684   6,927,124  ENSSSCG00000013035 PPP1R14B  6,932,322   6,934,762  ENSSSCG00000013036 FKBP2  6,935,189   6,936,674  ENSSSCG00000013037 VEGFB  6,939,770   6,942,952  ENSSSCG00000013038 DNAJC4  6,943,440   6,947,235  ENSSSCG00000013039 NUDT22  6,947,812   6,950,624  ENSSSCG00000013040 TRPT1  6,950,712   6,953,739  ENSSSCG00000013041 FERMT3  6,953,588   6,976,696  ENSSSCG00000021620   6,978,383   7,040,372  ENSSSCG00000013042 CCDC88B  6,992,733   7,010,294  ENSSSCG00000028806 RPS6KA4  7,012,544   7,025,897  ENSSSCG00000013043 MACROD1  7,054,730   7,204,542  ENSSSCG00000013044 FLRT1  7,095,830   7,097,924    75 Table S.4  ENSSSCG00000013045 OTUB1  7,204,713   7,212,472  ENSSSCG00000013046   7,218,448   7,220,218  ENSSSCG00000022687 NAA40  7,282,375   7,290,873  ENSSSCG00000013049 RCOR2  7,350,357   7,352,786  ENSSSCG00000013050 MARK2  7,355,161   7,364,707  ENSSSCG00000013048 C11orf84  7,464,512   7,506,772  ENSSSCG00000013051   7,516,650   7,517,021  ENSSSCG00000013053 C11orf95  7,550,065   7,555,127  ENSSSCG00000013052 RTN3  7,559,962   7,714,023  ENSSSCG00000013054   7,615,311   7,649,588  ENSSSCG00000026914 PLA2G16  7,653,511   7,681,456    76  Table S.5 List of genes in the SSC5 QTL region for TEN ordered according to increasing start position expressed in bp.  ENSEMBLE ID Gene Name Start position End Position ENSSSCG00000021596 KCNA5 67,653,414 67,655,144 ENSSSCG00000000716 KCNA1 67,768,663 67,770,150 ENSSSCG00000000717  67,863,129 67,864,757 ENSSSCG00000000718 GALNT8 67,903,769 67,941,362 ENSSSCG00000000719  67,965,977 67,994,531 ENSSSCG00000000720 AKAP3 68,003,279 68,020,465 ENSSSCG00000000721 DYRK4 68,022,319 68,047,436 ENSSSCG00000000722 RAD51AP1 68,079,112 68,103,087 ENSSSCG00000000723 C12orf4 68,104,334 68,145,457 ENSSSCG00000000724 FGF6 68,198,018 68,207,512 ENSSSCG00000029028 FGF23 68,250,561 68,261,760 ENSSSCG00000024219 TIGAR 68,266,578 68,280,421 ENSSSCG00000000727 CCND2 68,314,898 68,331,943 ENSSSCG00000000728 PARP11 68,750,885 68,791,407 ENSSSCG00000000730 PRMT8 68,942,252 69,061,177 ENSSSCG00000000732 CRACR2A 69,023,638 69,052,093   77  Table S.6 List of genes in the SSC6 QTL region for BF10, LLBF and LW ordered according to increasing start position expressed in bp.  ENSEMBLE ID Gene Name Start position End Position ENSSSCG00000003777 SLC44A5  127,567,588   127,970,326  ENSSSCG00000020641 U6  127,659,262   127,659,367  ENSSSCG00000003778 LHX8  128,034,488   128,059,465  ENSSSCG00000003779 TYW3  128,286,279   128,301,919  ENSSSCG00000003780 CRYZ  128,304,889   128,327,585  ENSSSCG00000003781 ERICH3  128,400,155   128,476,446  ENSSSCG00000003782  128,673,923   128,673,994  ENSSSCG00000003783 FPGT  128,715,728   128,722,791  ENSSSCG00000003784 LRRIQ3  128,721,115   128,766,182  ENSSSCG00000020493 5S_rRNA  129,093,391   129,093,527  ENSSSCG00000025085 NEGR1  130,680,416   130,880,673  ENSSSCG00000003787 ZRANB2  131,542,884   131,563,408  ENSSSCG00000019065 ssc-mir-186  131,558,966   131,559,047  ENSSSCG00000003788 PTGER3  131,574,481   131,616,324  ENSSSCG00000003789 CTH  132,027,412   132,050,286  ENSSSCG00000003790 ANKRD13C  132,080,136   132,189,620  ENSSSCG00000003791 SRSF11  132,193,667   132,240,405  ENSSSCG00000003792 LRRC40  132,240,462   132,288,460  ENSSSCG00000003793 LRRC7  132,309,803   132,436,579  ENSSSCG00000023754  132,544,632   132,612,363  ENSSSCG00000003794 RPE65  133,492,012   133,513,894  ENSSSCG00000003795 WLS  133,752,325   133,820,951  ENSSSCG00000003797 DIRAS3  133,865,007   133,867,486  ENSSSCG00000003798 SERBP1  134,068,990   134,081,998  ENSSSCG00000003799 IL12RB2  134,096,226   134,161,817  ENSSSCG00000003800  134,197,114   134,197,933  ENSSSCG00000003801 IL23R  134,219,683   134,283,657  ENSSSCG00000003802 SLC35D1  134,374,824   134,382,782  ENSSSCG00000003803 C1orf141  134,410,380   134,464,826  ENSSSCG00000030423 MIER1  134,486,591   134,497,982  ENSSSCG00000023741  134,537,789   134,545,301  ENSSSCG00000003805 PDE4B  134,870,923   134,913,618    78 Table S.6  ENSSSCG00000003806 LEPROT  135,379,661   135,387,507  ENSSSCG00000003807 DNAJC6  135,397,188   135,560,153  ENSSSCG00000003808  135,587,479   135,663,171  ENSSSCG00000018551 5S_rRNA  135,686,780   135,686,910  ENSSSCG00000019990 ssc-mir-101-2  135,736,114   135,736,196  ENSSSCG00000019115  135,812,759   135,812,840  ENSSSCG00000003809 JAK1  135,899,356   135,917,892  ENSSSCG00000025672 RAVER2  135,984,034   136,062,365  ENSSSCG00000004829 CACHD1  136,116,938   136,181,931  ENSSSCG00000003810 UBE2U  136,685,425   136,756,970  ENSSSCG00000003811 ROR1  136,765,918   136,909,446  ENSSSCG00000028540 U6  136,874,221   136,874,327  ENSSSCG00000003812 PGM1  137,171,271   137,233,581  ENSSSCG00000003814 EFCAB7  137,259,293   137,316,747  ENSSSCG00000020427 SNORA16  137,320,904   137,321,016  ENSSSCG00000003815 ALG6  137,377,946   137,434,247  ENSSSCG00000003816 ATG4C  137,803,101   137,886,719  ENSSSCG00000030607 U6  138,028,937   138,029,043  ENSSSCG00000003818 DOCK7  138,042,658   138,172,957  ENSSSCG00000003819 ANGPTL3  138,101,691   138,111,290  ENSSSCG00000003821 KANK4  138,309,636   138,383,020  ENSSSCG00000023243  138,764,544   138,851,088  ENSSSCG00000003823 C1orf87  139,665,483   139,746,337  ENSSSCG00000003824  139,867,650   139,892,000  ENSSSCG00000021940 CYP2J34  139,902,194   139,929,805  ENSSSCG00000023035 U6  139,902,298   139,902,404  ENSSSCG00000003825 CYP2J2  139,936,092   139,965,989  ENSSSCG00000019515 U6  139,946,008   139,946,112  ENSSSCG00000003826 HOOK1  140,117,181   140,190,970  ENSSSCG00000022526  140,248,872   140,253,789  ENSSSCG00000003827  140,363,964   140,393,806  ENSSSCG00000003828  140,421,878   140,676,185      79  Table S.7 List of genes in the SSC7 QTL region associated with loin muscle area and dressing percentage, ordered according to increasing start position expressed in bp.  ENSEMBLE ID Gene Name Start position  End Position  ENSSSCG00000001493  32750198 33022968 ENSSSCG00000019940 U6 32784377 32784487 ENSSSCG00000029539 RAB23 33091878 33119792 ENSSSCG00000001494 BAG2 33129268 33140672 ENSSSCG00000026207  33140918 33141047 ENSSSCG00000001497 ZNF451 33144581 33226714 ENSSSCG00000001498 BEND6 33231047 33273159 ENSSSCG00000025172  33274353 33374712 ENSSSCG00000001499 DST 33504832 33751812 ENSSSCG00000020421 SNORA72 33632171 33632302 ENSSSCG00000001500 COL21A1 33918801 34009970 ENSSSCG00000001501 VPS52 34128551 34142395 ENSSSCG00000001502 RPS18 34142621 34146677 ENSSSCG00000001503 B3GALT4 34146955 34148615 ENSSSCG00000001504 WDR46 34148893 34157078 ENSSSCG00000001505 PFDN6 34157443 34158709 ENSSSCG00000001506 RGL2 34159507 34166143 ENSSSCG00000001507 TAPBP 34167701 34176411 ENSSSCG00000001508 ZBTB22 34177375 34180363 ENSSSCG00000001509 DAXX 34180626 34184970 ENSSSCG00000001510  34198532 34213249 ENSSSCG00000001517  34273783 34278594 ENSSSCG00000001516 BAK1 34279881 34288116 ENSSSCG00000001515 ZBTB9 34338560 34339966 ENSSSCG00000001513 SYNGAP1 34342914 34371345 ENSSSCG00000001512 CUTA 34373453 34375042 ENSSSCG00000001511 PHF1 34375137 34380476 ENSSSCG00000028301  34381247 34387660 ENSSSCG00000001518 ITPR3 34443056 34510838 ENSSSCG00000030668 UQCC2 34506319 34537625 ENSSSCG00000001520 LEMD2 34641299 34656755 ENSSSCG00000001521 MLN 34663313 34672854 ENSSSCG00000001523 GRM4 34839241 34928071 ENSSSCG00000001526 HMGA1 34984737 34990089   80 Table S.7  ENSSSCG00000023160  34990224 34992729 ENSSSCG00000001528 RPS10 35109882 35117436 ENSSSCG00000024531 SPDEF 35214365 35233219 ENSSSCG00000027053 PACSIN1 35236544 35244143 ENSSSCG00000001527 C6orf106 35362835 35410240 ENSSSCG00000001531 SNRPC 35451546 35466721 ENSSSCG00000001532 UHRF1BP1 35474279 35538654 ENSSSCG00000001533 TAF11 35543524 35555308 ENSSSCG00000001534 ANKS1A 35556631 35766117 ENSSSCG00000001535 TCP11 35783270 35807767 ENSSSCG00000027696 U6 35795518 35795624 ENSSSCG00000001536 SCUBE3 35854337 36008752 ENSSSCG00000001544 TEAD3 36085701 36100686 ENSSSCG00000001543  36106602 36108831 ENSSSCG00000001546 FANCE 36111076 36123998 ENSSSCG00000025377  36127213 36135077 ENSSSCG00000001539 PPARD 36141606 36215260 ENSSSCG00000001538 DEF6 36220769 36243212 ENSSSCG00000001537 ZNF76 36245632 36254505 ENSSSCG00000001549  36373069 36478480 ENSSSCG00000001550 ARMC12 36518830 36529646 ENSSSCG00000001551 CLPSL2 36533600 36535664 ENSSSCG00000001552 CLPS 36546797 36549010 ENSSSCG00000001553 LHFPL5 36552501 36560440 ENSSSCG00000001554 SRPK1 36575699 36638641 ENSSSCG00000022412  36597506 36600159 ENSSSCG00000001555 SLC26A8 36660474 36714021 ENSSSCG00000001556 MAPK14 36725707 36795310 ENSSSCG00000020803  36804604 36865913 ENSSSCG00000001559 PNPLA1 36888272 36925502 ENSSSCG00000001560 C6orf222 36937623 36949206 ENSSSCG00000001561 ETV7 36990013 37008146 ENSSSCG00000001562 KCTD20 37046645 37093046 ENSSSCG00000001563 STK38 37106646 37154544 ENSSSCG00000001564  37191603 37198743 ENSSSCG00000022111  37258053 37267163 ENSSSCG00000001566 RAB44 37291881 37305207 ENSSSCG00000001567 CPNE5 37314126 37411217   81 Table S.7  ENSSSCG00000030390 PPIL1 37424434 37448185 ENSSSCG00000001569 C6orf89 37451398 37483062 ENSSSCG00000001570 PI16 37504337 37514234 ENSSSCG00000001571 MTCH1 37520856 37539035 ENSSSCG00000001572 FGD2 37551887 37578005 ENSSSCG00000001573 PIM1 37691963 37697467 ENSSSCG00000001574  37721577 37722092 ENSSSCG00000001575  37724093 37724653 ENSSSCG00000001576 TMEM217 37728884 37729462 ENSSSCG00000001577  37742829 37818510 ENSSSCG00000001578 RNF8 37836027 37867217 ENSSSCG00000028436 SNORA70 37891821 37891951 ENSSSCG00000001579  37905566 37968899 ENSSSCG00000027292  37970118 37990894 ENSSSCG00000001582 MDGA1 38114185 38140436 ENSSSCG00000001583  38173326 38173520 ENSSSCG00000001584  38746485 38878917 ENSSSCG00000021215 ZFAND3 38909262 38909459 ENSSSCG00000030073  38947185 39089845 ENSSSCG00000027778 GLO1 39241001 39271606 ENSSSCG00000001588 DNAH8 39286129 39569290 ENSSSCG00000001589 GLP1R 39595053 39623436 ENSSSCG00000025423 KCNK5 39704035 39743285 ENSSSCG00000001592 KCNK17 39796871 39811137    82 Table S.8 List of genes in the SSC7 QTL region associated with number of ribs and carcass length, ordered according to increasing start position expressed in bp.  ENSEMBLE ID Gene Name Start position  End Position  ENSSSCG00000022178 102416956 102504747 ENSSSCG00000002344 102610127 102612130 ENSSSCG00000002345 102665568 102680830 ENSSSCG00000002346 PNMA1 102682182 102683240 ENSSSCG00000002347 DNAL1 102691235 102733635 ENSSSCG00000002348 ACOT6 102742330 102748593 ENSSSCG00000030559 102770509 102771156 ENSSSCG00000002349 ACOT4 102786805 102791429 ENSSSCG00000002350 102828131 102831810 ENSSSCG00000020600 102900731 102900857 ENSSSCG00000002351 102928581 102950219 ENSSSCG00000002352 ZNF410 102952601 103001300 ENSSSCG00000002353 FAM161B 102997112 103012716 ENSSSCG00000002354 COQ6 103012820 103025093 ENSSSCG00000002355 ENTPD5 103028636 103046142 ENSSSCG00000002356 BBOF1 103070839 103114503 ENSSSCG00000002357 ALDH6A1 103117991 103136318 ENSSSCG00000027564 LIN52 103136846 103337802 ENSSSCG00000024942 103192746 103202335 ENSSSCG00000002359 103231353 103248739 ENSSSCG00000002363 VSX2 103375383 103394310 ENSSSCG00000002362 103411781 103428283 ENSSSCG00000002361 VRTN 103457506 103467076 ENSSSCG00000025000 SYNDIG1L 103502244 103522505 ENSSSCG00000002366 NPC2 103573074 103582623 ENSSSCG00000002367 ISCA2 103582741 103584289 ENSSSCG00000002368 LTBP2 103590368 103694062 ENSSSCG00000002370 AREL1 103736066 103783771 ENSSSCG00000021442 FCF1 103783793 103805586 ENSSSCG00000002372 103818763 103886143 ENSSSCG00000002373 PROX2 103908499 103918733 ENSSSCG00000002373 PROX2 103908499 103918733 ENSSSCG00000002374 DLST 103934687 103955706 ENSSSCG00000002375 RPS6KL1 103957391 103974114   83 Table S.8  ENSSSCG00000002376 PGF 103994328 104006524 ENSSSCG00000002376 PGF 103994328 104006524 ENSSSCG00000002377 EIF2B2 104047923 104054504 ENSSSCG00000002378 MLH3 104064666 104089446 ENSSSCG00000002379 ACYP1 104093822 104109273 ENSSSCG00000002380 ZC2HC1C 104110332 104117479 ENSSSCG00000002381 NEK9 104120031 104166205 ENSSSCG00000002382 TMED10 104173906 104206702 ENSSSCG00000002383 FOS 104293657 104297121 ENSSSCG00000021606 104424137 104479243 ENSSSCG00000030582 BATF 104503137 104524774 ENSSSCG00000024255 104608351 104625059 ENSSSCG00000002384 104685217 104811317 ENSSSCG00000029832 7SK 104854284 104854566 ENSSSCG00000025984 TTLL5 104990613 105175649 ENSSSCG00000025984 TTLL5 104990613 105175649 ENSSSCG00000029713 7SK 105034004 105034286 ENSSSCG00000002385 TGFB3 105178633 105206982 ENSSSCG00000002385 TGFB3 105178633 105206982 ENSSSCG00000002386 IFT43 105268438 105299304 ENSSSCG00000002387 GPATCH2L 105370887 105416373 ENSSSCG00000002388 ESRRB 105628140 105743716 ENSSSCG00000018299 105886390 105886534 ENSSSCG00000002389 VASH1 105960537 105978240 ENSSSCG00000002390 ANGEL1 105992708 105999604 ENSSSCG00000002391 LRRC74A 106005464 106033717 ENSSSCG00000002392 IRF2BPL 106151786 106154137    84 Table S.9 List of genes in the SSC15 QTL region for juiciness, tenderness, WBS, 24-h pH, protein content, CY and drip loss ordered according to increasing start position expressed in bp.  ENSEMBLE ID Gene Name Start position End Position ENSSSCG00000020722 SNORA42  133,359,013   133,359,150  ENSSSCG00000029002  133,367,666   133,370,494  ENSSSCG00000016185 ARPC2  133,377,438   133,389,903  ENSSSCG00000021228 SNORA42  133,415,454   133,415,591  ENSSSCG00000016186  133,425,398   133,432,043  ENSSSCG00000022483  133,442,237   133,443,117  ENSSSCG00000016189  133,444,024   133,447,622  ENSSSCG00000025058  133,452,329   133,456,736  ENSSSCG00000016187  133,462,315   133,468,029  ENSSSCG00000016196 VIL1  133,479,159   133,507,423  ENSSSCG00000016194 USP37  133,512,791   133,595,116  ENSSSCG00000016195  133,583,661   133,584,166  ENSSSCG00000016193 RQCD1  133,604,447   133,636,861  ENSSSCG00000016192  133,641,605   133,669,136  ENSSSCG00000016191  133,667,702   133,729,988  ENSSSCG00000026964  133,753,315   133,852,318  ENSSSCG00000016201 TTLL11  133,768,369   133,815,768  ENSSSCG00000016200 PRKAG3  133,800,248   133,807,019  ENSSSCG00000028455 TTLL4  133,817,190   133,818,795  ENSSSCG00000016199  133,828,770   133,831,213  ENSSSCG00000016202  133,834,386   133,837,207  ENSSSCG00000023145  133,855,102   133,856,393  ENSSSCG00000021584 CDK5R2  133,925,849   133,927,094  ENSSSCG00000026963  133,957,554   133,957,623  ENSSSCG00000026958  134,033,487   134,033,948  ENSSSCG00000016203 CCDC108  134,087,004   134,103,779  ENSSSCG00000016204 IHH  134,122,695   134,129,391  ENSSSCG00000016205 NHEJ1  134,145,208   134,232,523  ENSSSCG00000016213 GLB1L  134,233,624   134,318,259  ENSSSCG00000029694 SLC23A3  134,234,854   134,240,678  ENSSSCG00000016206 CNPPD1  134,243,394   134,248,180  ENSSSCG00000016207 FAM134A  134,249,648   134,255,270  ENSSSCG00000016208 ZFAND2B  134,277,911   134,280,873  ENSSSCG00000016210 ABCB6  134,280,965   134,288,809    85 Table S.9  ENSSSCG00000016211 ATG9A  134,289,150   134,300,597  ENSSSCG00000018394  134,294,632   134,294,714  ENSSSCG00000016212 ANKZF1  134,300,649   134,309,215  ENSSSCG00000016214 STK16  134,320,058   134,323,470  ENSSSCG00000016215 TUBA4A  134,325,130   134,328,954  ENSSSCG00000016216  134,341,043   134,347,416  ENSSSCG00000016217 DNAJB2  134,355,300   134,360,929  ENSSSCG00000016218 PTPRN  134,364,671   134,384,701  ENSSSCG00000018628  134,369,134   134,369,275  ENSSSCG00000016219 RESP18  134,398,121   134,404,728  ENSSSCG00000022460  134,407,248   134,407,824  ENSSSCG00000016220 DNPEP  134,425,137   134,437,835  ENSSSCG00000030434 ssc-mir-4334  134,433,005   134,433,073  ENSSSCG00000020771 INHA  134,507,304   134,511,349  ENSSSCG00000028052 OBSL1  134,512,277   134,533,002  ENSSSCG00000027541 TMEM198  134,534,143   134,587,352  ENSSSCG00000020785 DES  134,560,460   134,567,338  ENSSSCG00000021610 CHPF  134,588,465   134,594,046  ENSSSCG00000029968 ASIC4  134,594,938   134,619,791  ENSSSCG00000023935 GMPPA  134,627,636   134,635,226  ENSSSCG00000018784 U6  135,384,999   135,385,103  ENSSSCG00000020038 SNORA31  135,665,635   135,665,753  ENSSSCG00000026566  135,772,108   135,772,198  ENSSSCG00000016230 EPHA4  136,746,506   136,874,945  ENSSSCG00000016231  136,930,815   136,936,376      86           LITERATURE CITED  87 LITERATURE CITED   Akanno, E. C., Plastow, G., B. W. Woodwards, S. Bauck, H. Okut, X.-L. Wu, C. Sun, J. L. Aalhus, S. S. Moore, S. P. Miller, W. Z., and J. A. Basarab. 2014. Reliability of molecular breeding values for Warner-Bratzler shear force and carcass traits of beef cattle - an independent validation study. J. Anim. Sci. 92:28962904. http://doi.org/10.2527/jas.2013-7374 Badke, Y. M., R. O. Bates, C. W. Ernst, C. Schwab, J. Fix, C. P. Van Tassell, and J. P. Steibel. 2013. Methods of tagSNP selection and other variables affecting imputation accuracy in swine. BMC Genet. 14:8. http://doi.org/10.1186/1471-2156-14-8 Becker, D., K. Wimmers, H. Luther, A. Hofer, and T. Leeb. 2013. A Genome-Wide Association Study to Detect QTL for Commercially Important Traits in Swiss Large White Boars. PLoS One 8:16. http://doi.org/10.1371/journal.pone.0055951 Benjamini, Y., and Y. Hochberg. 1995. ContrPowerful Approach to Multiple Testing When researchers tend to select pursuing multiple the ( statistically ) and support of conclusions . An unguarded use in a greatly results of single-inference inc. J.R Stat. Soc.B 57:289300. http://doi.org/10.2307/234610 Berg, E. P., E. L. McFadin, K. R. Maddock, R. N. Goodwin, T. J. Baas, and D. H. Keisler. 2003. Serum concentrations of leptin in six genetic lines of swine and relationship with growth and carcass characteristics. J. Anim. Sci. 81:167171. Bernal Rubio, Y. L., J. L. Gualdrón Duarte, R. O. Bates, C. W. Ernst, D. Nonneman, G. A. Rohrer, A. King, S. D. Shackelford, T. L. Wheeler, R. J. C. Cantet, and J. P. Steibel. 2015a. Implementing meta-analysis from genome-wide association studies for pork quality traits 1. J. Anim. Sci. 93:56075617. http://doi.org/10.2527/jas2015-950 Bernal Rubio, Y. L., J. L. Gualdrón Duarte, R. O. Bates, C. W. Ernst, D. Nonneman, G. A. Rohrer, A. King, S. D. Shackelford, T. L. Wheeler, R. J. C. Cantet, and J. P. Steibel. 2015b. Meta-analysis of genome-wide association from genomic prediction models. Anim. Genet.:3648. http://doi.org/10.1111/age.12378 Chen, C. Y., I. Misztal, I. Aguilar, S. Tsuruta,  the Meuwissen, S. E. Aggrey, T. Wing, and W. M. Muir. 2011. Genome-wide marker-assisted selection combining all pedigree phenotypic information with genotyping. J. Anim. Sci. 89:23. http://doi.org/10.2527/jas.2010-3071 Choi, I., J. P. Steibel, R. O. Bates, N. E. Raney, J. M. Rumph, and C. W. Ernst. 2011. Population Using Different Least-Squares Analysis Models. Front. Genet. 2:18. http://doi.org/10.3389/fgene.2011.00018 Ciobanu, D., J. Bastiaansen, M. Malek, J. Helm, J. Woollard, G. Plastow, and M. Rothschild. 2001. Evidence for New Alleles in the Protein Kinase Adenosine Monophosphate- Activated  3 -Subunit Gene Associated With Low Glycogen Content in Pig Skeletal Muscle and Improved Meat Quality. Genetics 159:11511162.   88 Couturier, C., C. Sarkis, K. Séron, S. Belouzard, P. Chen, A. Lenain, L. Corset, J. Dam, V. Vauthier, A. Dubart, J. Mallet, P. Froguel, Y. Rouillé, and R. Jockers. 2007. Silencing of OB-RGRP in mouse hypothalamic arcuate nucleus increases leptin receptor signaling and prevents diet-induced obesity. Proc. Natl. Acad. Sci. U. S. A. 104:1947619481. http://doi.org/10.1073/pnas.0706671104 da Costa,  a. S. H., V. M. R. Pires, C. M. G. a. Fontes, and J. a. M. Prates. 2013. Expression of genes controlling fat deposition in two genetically diverse beef cattle breeds fed high or low silage diets. BMC Vet. Res. 9:118. http://doi.org/10.1186/1746-6148-9-118 Edwards, D. B., C. W. Ernst, N. E. Raney, M. . Doumit, M. D. Hoge, and R. O. Bates. 2008. Quantitative trait loci mapping in an F2 Duroc x Pietrain resource population: II. Carcass and meat quality traits. J. Anim. Sci. 86:254:266. http://doi.org/10.2527/jas.2006-626 Edwards, D., R. Bates, and W. Osburn. 2003. Evaluation of Duroc-vs. Pietrain-sired pigs for carcass and meat quality measures. J. Anim. Sci.:18951899. Ernst, C. W., P. A. Kapke, M. Yerle, and M. F. Rothschild. 1996. The leptin receptor gene (LEPR) maps to porcine chromosome 6. Mamalian genome 8:226. Fan, Y., Y. Xing, Z. Zhang, H. Ai, Z. Ouyang, J. Ouyang, M. Yang, P. Li, Y. Chen, J. Gao, L. Li, L. Huang, and J. Ren. 2013. A further look at porcine chromosome 7 reveals VRTN variants associated with vertebral number in Chinese and Western pigs. PLoS One 8:e62534. http://doi.org/10.1371/journal.pone.0062534 Forneris, N. S., A. Legarra, Z. G. Vitezica, S. Tsuruta, I. Aguilar, I. Misztal, and R. J. C. Cantet. 2015. Quality control of genotypes using heritability estimates of gene content at the marker. Genetics 199:67581. http://doi.org/10.1534/genetics.114.173559 Geldermann, H., E. Muller, G. Moser, G. Reiner, H. Bartenschlager, S. Cepica, A. Stratil, J. Kuryl, C. Moran, R. Davoli, and C. Brunsch. 2003. Genome-wide linkage and QTL mapping in porcine F2 families generated from Pietrain, Meishan and Wild Boar crosses. J. Anim. Breed. Genet. 120:363393. http://doi.org/10.1046/j.0931-2668.2003.00408 Goddard, M. E., and B. J. Hayes. 2007. Genomic selection. J.Anim.Breed.Genet 124:323330. http://doi.org/10.1080/09064700801959395 Goll, D. E., V. F. Thompson, R. G. Taylor, and T. Zalewska. 1992. Is calpain activity regulated by membranes and autolysis or by calcium and calpastatin? Bioessays 14:54956. http://doi.org/10.1002/bies.950140810 Gualdrón Duarte, J. L., R. O. Bates, C. W. Ernst, N. E. Raney, R. J. C. Cantet, and J. P. Steibel. 2013. Genotype imputation accuracy in a F2 pig population using high density and low density SNP panels. BMC Genet. 14:38. http://doi.org/10.1186/1471-2156-14-38 Gualdrón Duarte, J. L., R. J. C. Cantet, R. O. Bates, C. W. Ernst, N. E. Raney, and J. P. Steibel. 2014. Rapid screening for phenotype-genotype associations by linear transformations of genomic evaluations. BMC Bioinformatics 15:246. http://doi.org/10.1186/1471-2105-15-246 Gualdrón Duarte, J. L.; Cantet, R. J. C.; Bernal Rubio, Y. L.; Bates, R. O.; Ernst, C. W.; Raney,   89 N. E.; Rogberg-Muñoz, A.; Steibel, J. P. 2016. Refining genome wide association for growth and fat deposition traits in an F2 pig population1. Journal of Animal Science 94 http://doi.org/10.2527/jas2015-0182 Hayes, B. J., P. J. Bowman, A. J. Chamberlain, and M. E. Goddard. 2009. Genomic selection in dairy cattle: progress and challenges. J. Dairy Sci. 92:433443. http://doi.org/10.3168/jds.2008-1646 Hayes, B. J., J. Pryce, A. J. Chamberlain, P. J. Bowman, and M. E. Goddard. 2010. Genetic architecture of complex traits and accuracy of genomic Prediction: Coat colour, Milk-fat percentage, and type in holstein cattle as contrasting model traits. PLoS Genet. 6. http://doi.org/10.1371/journal.pgen.1001139 Hayes, B. J. 2013. Overview of statistical methods for genome-wide association studies. C. Gondor, B. van der Werf, & B. J. Hayes, editors, Genome-wide association studies and genomic prediction. Human Press, New York, NY. p. 156-157. Hu, Z.-L., C. A. Park, and J. M. Reecy. 2015. Developmental progress and current status of the Animal QTLdb. Nucleic Acids Res. 44:D827833. http://doi.org/10.1093/nar/gkv1233 Huff-Lonergan, E., T. Mitsuhashi, D. D. Beekman, F. C. Parrish, D. G. Olson, and R. M. -Calpain at Low pH and Temperature is Similar to Degradation in Postmortem Bovine Muscle. J. Anim. Sci. 74:9931008. http://doi.org//1996.745993x Janss, L., G. de los Campos, N. Sheehan, and D. Sorensen. 2012. Inferences from genomic models in stratified populations. Genetics 192:693704. http://doi.org/10.1534/genetics.112.141143 Jogl, G., Y. S. Hsiao, and L. Tong. 2004. Structure and function of carnitine acyltransferases. Ann. N. Y. Acad. Sci. 1033:1729. http://doi.org/10.1196/annals.1320.002 Kim, E. S., R. Ros-Freixedes, R. N. Pena, T. J. Bass, J. Estany, and M. F. Rothschild. 2015. Identification of signatures of selection for intramuscular fat and backfat thickness in teo Duroc populations. J. Anim. Sci. 93:32933302. http://doi.org/10.2527/jas2015-8879 Kim, J. J., M. F. Rothschild, J. Beever, S. Rodriguez-Zas, and J. C. M. Dekkers. 2005. Joint analysis of two breed cross populations in pigs to improve detection and characterization of quantitative trait loci. J. Anim. Sci. 83:12291240. http://doi.org/83/6/1229 King, J. W. B., and R. C. Roberts. 1960. Carcass length in the bacon pig; its association with vertebrae numbers and prediction from radiographs of the young pig. Anim. Prod. 2:5965. http://doi.org/10.1017/S0003356100033493 Koohmaraie, M. 1992. Ovine skeletal muscle multicatalytic proteinase complex (proteasome): purification, characterization, and comparison of its effects on myofibrils with mu-calpains. J. Anim. Sci. 70:36973708. http://doi.org/1474009 Kühn, K. 1995. Basement membrane (type IV) collagen. Matrix Biol. 14:439445.   90 Li, H. D., M. S. Lund, O. F. Christensen, V. R. Gregersen, P. Henckel, and C. Bendixen. 2010. Quantitative trait loci analysis of swine meat quality traits. J. Anim. Sci. 88:29042912. http://doi.org/10.2527/jas.2009-2590 Liu, G., D. G. J. Jennen, E. Tholen, H. Juengst, T. Kleinwächter, M. Hölker, D. Tesfaye, G. Un, H.-J. Schreinemachers, E. Murani, S. Ponsuksili, J.-J. Kim, K. Schellander, and K. Wimmers. 2007. A genome scan reveals QTL for growth, fatness, leanness and meat quality in a Duroc-Pietrain resource population. Anim. Genet. 38:24152. http://doi.org/10.1111/j.1365-2052.2007.01592.x Lonergan, S. M., E. Huff-Lonergan, L. J. Rowe, D. L. Kuhlers, and S. B. Jungst. 2001. Selection for lean growth efficiency in Duroc pigs influences pork quality. J. Anim. Sci. 79:20752085. http://doi.org//2001.7982075x Ma, J., J. Yang, L. Zhou, Z. Zhang, H. Ma, X. Xie, F. Zhang, X. Xiong, L. Cui, H. Yang, X. Liu, Y. Duan, S. Xiao, H. Ai, J. Ren, and L. Huang. 2013. Genome-wide association study of meat quality traits in a White Duroc×Erhualian F2 intercross and Chinese Sutai pigs. PLoS One 8:e64047. http://doi.org/10.1371/journal.pone.0064047 Malek, M., J. C. M. Dekkers, H. K. Lee, T. J. Baas, and M. F. Rothschild. 2001. A molecular genome scan analysis to identify chromosomal regions influencing economic traits in the pig. I. Growth and body composition. Mamm. Genome 12:630636. http://doi.org/10.1007/s003350020018 Meuwissen, T. H. E., B. J. Hayes, and M. E. Goddard. 2001. Prediction of total genetic value using genome-wide dense marker maps. Genetics 157:18191829. http://doi.org/11290733 Milan, D., J. P. Bidanel, N. Iannuccelli, J. Riquet, Y. Amigues, J. Gruand, P. Le Roy, C. Renard, and C. Chevalet. 2002. Detection of quantitative trait loci for carcass composition traits in pigs. Genet. Sel. Evol. 34:705728. http://doi.org/10.1051/gse Milan, D., J. T. Jeon, C. Looft, V. Amarger, A. Robic, M. Thelander, C. Rogel-Gillard, S. Paul, N. Iannuccelli, L. Rask, H. Ronne, K. Lundström, N. Reinsch, J. Gellin, E. Kalm, P. Le Roy, P. Chardon, and L. Andersson. 2000. A Mutation in PRKAG3 Associated with Excess Glycogen Content in Pig Skeletal Muscle. Science (80-. ). 288:12481251. http://doi.org/10.1126/science.288.5469.1248 Muñoz, G., E. Alcazar, A. Fernández, C. Barragán, A. Carrasco, E. de Pedro, L. Silio, J. L. Sanchez, and M. C. Rodriguez. 2011. Effects of porcine MC4R and LEPR polymorphisms, gender and Duroc sire line on economic traits in Duroc??Iberian crossbred pigs. Meat Sci. 88:169173. http://doi.org/10.1016/j.meatsci.2010.12.018 Muñoz, G., C. Ovilo, L. Silló, A. Tomás, J. L. Noguera, and M. C. Rodríguez. 2009. Single- And joint-population analyses of two experimental pig crosses to confirm quantitative trait loci on Sus scrofa chromosome 6 and leptin receptor effects on fatness and growth traits. J. Anim. Sci. 87:459468. http://doi.org/10.2527/jas.2008-1127 Nakano, H., S. Sato, Y. Uemoto, T. Kikuchi, T. Shibata, H. Kadowaki, E. Kobayashi, and K. Suzuki. 2015. Effect of VRTN gene polymorphisms on Duroc pig production and carcass   91 traits, and their genetic relationships. Anim. Sci. J. 86:125131. http://doi.org/10.1111/asj.12260 Nonneman, D. J., S. D. Shackelford, D. A. King, T. L. Wheeler, R. T. Wiedmann, W. M. Snelling, and G. A. Rohrer. 2013. Genome-wide association of meat quality traits and tenderness in swine D. J. Nonneman, S. D. Shackelford, D. A. King, T. L. Wheeler, R. T. Wiedmann, W. M. Snelling and G. A. Rohrer. J. Anim. Sci.:40434050. http://doi.org/10.2527/jas2013-6255 Ovilo, C., A. Fernández, J. L. Noguera, C. Barragán, R. Letón, C. Rodríguez, A. Mercadé, E. Alves, J. M. Folch, L. Varona, and M. Toro. 2005. Fine mapping of porcine chromosome 6 QTL and LEPR effects on body composition in multiple generations of an Iberian by Landrace intercross. Genet. Res. 85:5767. http://doi.org/10.1017/S0016672305007330 Patel, K. G., C. Liu, P. L. Cameron, and R. S. Cameron. 2001. Myr 8, a novel unconventional myosin expressed during brain development associates with the protein phosphatase catalytic subunits 1alpha and 1gamma1. J. Neurosci. 21:79547968. Ramos, A. M., R. P. M. A. Crooijmans, N. A. Affara, A. J. Amaral, A. L. Archibald, J. E. Beever, C. Bendixen, C. Churcher, R. Clark, P. Dehais, M. S. Hansen, J. Hedegaard, Z.-L. Hu, H. H. Kerstens, A. S. Law, H.-J. Megens, D. Milan, D. J. Nonneman, G. A. Rohrer, M. F. Rothschild, T. P. L. Smith, R. D. Schnabel, C. P. Van Tassell, J. F. Taylor, R. T. Wiedmann, L. B. Schook, and M. A. M. Groenen. 2009. Design of a High Density SNP Genotyping Assay in the Pig Using SNPs Identified and Characterized by Next Generation Sequencing Technology. PLoS One 4:e6524. http://doi.org/10.1371/journal.pone.0006524 Rohrer, G. a, and J. W. Keelen. 1998. Identification of quantitative trait loci affecting Carcass Composition in swine: I. Fat Deposition Traits. J. Anim. Sci. 76:22472254. Rohrer, G. A., D. J. Nonneman, R. K. Miller, H. Zerby, and S. J. Moeller. 2012. Association of single nucleotide polymorphism (SNP) markers in candidate genes and QTL regions with pork quality traits in commercial pigs. Meat Sci. 92:511518. http://doi.org/10.1016/j.meatsci.2012.05.020 Rohrer, G. a., R. M. Thallman, S. Shackelford, T. Wheeler, and M. Koohmaraie. 2005. A genome scan for loci affecting pork quality in a Duroc-Landrace F2 population. Anim. Genet. 37:1727. http://doi.org/10.1111/j.1365-2052.2005.01368.x Ros-Freixedes, R., S. Gol, R. N. Pena, M. Tor, J. C. M. Dekkers, and J. Estany. 2014. Proceedings, 10. In: Genome-Wide Association Study for Intramuscular Fat content and Composition in Duroc Pigs. McGee, M. Gispert, O. I. Southwood, and T. Sweeney. 2012. SNP variation in the promoter of the PRKAG3 gene and association with meat quality traits in pig. BMC Genet. 13:66. http://doi.org/10.1186/1471-2156-13-66 Sanchez, M.-P., T. Tribout, N. Iannuccelli, M. Bouffaud, B. Servin, A. Tenghe, P. Dehais, N. Muller, M. P. Del Schneider, M.-J. Mercat, C. Rogel-Gaillard, D. Milan, J.-P. Bidanel, and   92 H. Gilbert. 2014. A genome-wide association study of production traits in a commercial population of Large White pigs: evidence of haplotypes affecting meat quality. Genet. Sel. Evol. 46:12. http://doi.org/10.1186/1297-9686-46-12 Sargolzaei, M., J. P. Chesnais, and F. S. Schenkel. 2014. A new approach for efficient genotype imputation using information from relatives. BMC Genomics 15:478. http://doi.org/10.1186/1471-2164-15-478 Sato, S., and Y. Oyamada. 2003. Quantitative trait loci analysis for growth and carcass traits in a 2949. Steibel, J. P., R. O. Bates, G. J. M. Rosa, R. J. Tempelman, V. D. Rilington, A. Ragavendran, N. E. Raney, A. M. Ramos, F. F. Cardoso, D. B. Edwards, and C. W. Ernst. 2011. Genome-wide linkage analysis of global gene expression in loin muscle tissue identifies candidate genes in pigs. PLoS One 6. http://doi.org/10.1371/journal.pone.0016766 Storey, J. D., and R. Tibshirani. 2003. Statistical significance for genomewide studies. 2003. Storey, J. D. 2002. A direct approach to false discovery rates. J. R. Stat. Soc. Series B. Stat. Methodol.:479498. http://doi.org/10.1111/1467-9868.00346 Tartaglia, L. a. 1997. The leptin receptor. J. Biol. Chem. 272:60936096. http://doi.org/10.1074/jbc.272.10.6093 Thomsen, H., H. K. Lee, M. F. Rothschild, M. Malek, and J. C. M. Dekkers. 2004. Characterization of quantitative trait loci for growth and meat quality in a cross between commercial breeds of swine 1. :22132228. Uemoto, Y., T. Kikuchi, H. Nakano, S. Sato, T. Shibata, H. Kadowaki, K. Katoh, E. Kobayashi, and K. Suzuki. 2012. Effects of porcine leptin receptor gene polymorphisms on backfat thickness, fat area ratios by image analysis, and serum leptin concentrations in a Duroc purebred population. Anim. Sci. J. 83:375385. http://doi.org/10.1111/j.1740-0929.2011.00963.x Uemoto, Y., and Y. Nagamine. 2008. Quantitative trait loci analysis on Sus scrofa chromosome 7 for meat production, meat quality, and carcass traits within a Duroc purebred population. J. -0293 Uimari, P., A. Sironen, and M. L. Sevón-Aimonen. 2013. Evidence for three highly significant QTL for meat quality traits in the Finnish Yorkshire pig breed. J. Anim. Sci. 91:20012011. http://doi.org/10.2527/jas.2012-5811 Uimari, P., and A. Sironen. 2014. A combination of two variants in PRKAG3 is needed for a positive effect on meat quality in pigs. BMC Genet. 15:29. http://doi.org/10.1186/1471-2156-15-29 VanRaden, P. M. 2008. Efficient Methods to Compute Genomic Predictions. J. Dairy Sci. 91:44144423. http://doi.org/10.3168/jds.2007-0980 Van Der Rest, M., and R. Garrone. 1991. Collagen family of proteins. FASEB J. 5:28142823.   93 Wang, H., I. Misztal, I. Aguilar, A. Legarra, and W. M. Muir. 2012. Genome-wide association mapping including phenotypes from relatives without genotypes. Genet. Res. (Camb). 94:7383. http://doi.org/10.1017/s0016672312000274 Wang, K., D. Liu, J. Hernandez-Sanchez, J. Chen, C. Liu, Z. Wu, M. Fang, and N. Li. 2015. Genome Wide Association Analysis Reveals New Production Trait Genes in a Male Duroc Population. PLoS One 10:e0139207. http://doi.org/10.1371/journal.pone.0139207 Wong, W., and J. . Scott. 2004. AKAP signalling complexes: focal points in space and time. Nat. Rev. 5:959970. Yoo, C.-K., I.-C. Cho, J.-B. Lee, E.-J. Jung, H.-T. Lim, S.-H. Han, S.-S. Lee, M.-S. Ko, T. Kang, J.-H. Hwang, Y. S. Park, and H.-B. Park. 2014. QTL analysis of body weight and carcass body length traits in an F2 intercross between Landrace and Korean native pigs. Anim. Genet. 45:589592. http://doi.org/10.1152/physiolgenomics.00172.2011 Zhang, C., Z. Wang, H. Bruce, R. A. Kemp, P. Charagu, Y. Miar, T. Yang, and G. Plastow. 2015. Genome-wide association studies (GWAS) identify a QTL close to PRKAG3 affecting meat pH and colour in crossbred commercial pigs. BMC Genet. 16:112. http://doi.org/10.1186/s12863-015-0192-1 Zhang, J. h., Y. Z. Xiong, B. Zuo, M. G. Lei, F. E. Li, and J. L. Li. 2007. Detection of Quantitative Trait Loci Associated with Live Measurement Traits in Pigs. Agric. Sci. China 6:863868. http://doi.org/10.1016/S1671-2927(07)60123-0    94 CHAPTER FOUR    95 Conclusion GOALS AND CONTRIBUTIONS OF THIS STUDY Meat quality traits are subject to genetic control, and are negatively correlated with lean growth traits (van Wijk et al., 2005). For a long time, the swine industry selected animals for lean growth and efficiency traits, inducing a softer and exudative meat (Lonergan et al., 2001). Additionally, using traditional breeding methods such as selection index (Hazel, 1943) is challenging for traits expressed later in life. With the development of statistical models for genomic selection (Meuwissen et al., 2001; Goddard & Hayes, 2007), genomic breeding values can be estimated, and animals can be selected earlier in life. However, genomic selection relies mostly on LD between markers and causative variants [such as Quantitative Trait Nucleotides (QTN)] as opposed to directly exploiting QTN (Weller & Ron, 2011). Detecting QTN has potential value in implementing selection across multiple population or for simultaneous multi-trait selection. More recently, the importance of knowing QTN has been highlighted by the prospect of using genome editing (Jenko et al., 2015). To detect a QTN, a Genome-Wide Association (GWA) is typically performed first, but in livestock sizable genomic regions are usually defined by GWA due to long range persistence of LD. Therefore, there is a need to refine the genomic regions using a method that can maximize the chances of covering the causal variants. The goals of this study were: 1. Implement and test properties of methods for computing the confidence interval of a QTL position in the context of GWA from mixed effects Genomic Best Linear Unbiased Predictor (GBLUP) models.    96 2. Perform GWA of meat quality and carcass traits in an F2 Duroc x Pietrain resource population using the methods tested under aim 1 and propose candidate genes for further study. In chapter 2, I used plasmode simulations to test the properties of three methods to compute Confidence Intervals (CI) for the position of Quantitative Trait Loci (QTL). Hayes, (2013) proposed a parametric method (PM) and I proposed two related non-parametric alternatives: one where the CI is centered around the QTL peak (NPC), and another one that produced asymmetric CI, non centered around the QTL peak (NPNC). I focused on two key properties of CI: Realized statistical coverage and length of the interval. The realized coverage of an optimal CI has to reach its nominal level using the shortest interval possible. The NPC failed to provide adequate coverage for the 95% CI for the QTL position. The NPNC and PM had very similar coverage, close to the nominal level (96% and 96.5% respectively). The asymmetry of the CI obtained with NPNC made this method adaptable to the density of significant SNP around the QTL peak at the expense of a longer CI length. None of the NPNC and PM were uniformly better than the other. In some cases, the 95% CI derived with the NPNC covered the true QTL position when the 95% CI derived with PM do not, and vice-versa. However, PM produced intervals on average 20% shorter than those from the NPNC. Therefore, I recommended to use the PM method for the calculation of CI for the QTL position in GWA and I used the method extensively in chapter 3. In chapter 3 I performed a GBLUP-based GWA in an F2 Duroc x Pietrain Resource Population for 38 meat quality and carcass traits and used statistical support intervals to map QTL. I found nine QTL associated with 15 traits on 8 chromosomes. Seven QTL had been previously reported. From those seven QTL, three (one on SSC1, tenth rib backfat thickness; one on SSC7, dressing percentage and loin muscle area; and one on SSC11, belly weight) had been previously mapped   97 only using low resolution linkage analyses. In this work, for the first time, those three QTL have been physically mapped to specific genomic segments. As a result, I proposed the gene Carnitine O-Acetyltransferase (CRAT; SSC1:303.4-303.41 Mb), which is related to lipid metabolism, as a candidate for tenth rib carcass backfat thickness. Two QTL associated with sensory panel tenderness on SSC3 and on SSC5 are novel findings of this study. The gene A kinase (PRKA) anchor protein 3 (AKAP3; SSC5: 68 to 68.02 Mb) is a novel candidate gene proposed for tenderness in this thesis. AKAP3 can bind to the regulatory subunit of PKA affecting the glycogen content in muscle, affecting the quality of the meat.  Finally, I studied an association peak for pH 24 hours post-mortem, drip loss and cook yield located close to the well-known candidate gene Protein Kinase AMP-activated 3-subunit (PRKAG3). My follow-up analysis focused on the association to two well-known non-synonymous variants (I199V and T30N) of PRKAG3 that have been proposed as candidate variants for pH 24 hours post mortem, drip loss and cook yield. In this thesis, I showed that in this population, those variants do not fully explain the genotypes of the QTL found on SSC15, associated with juiciness, tenderness, drip loss, pH 24 hours, WBS, cook yield and protein content. This means that the MSUPRP population remains a valuable resource to further discover more causative variants for pH 24 hours and related traits, either within PRKAG3 or in its close vicinity.  98 FUTURE RESEARCH DIRECTIONS: In this study I evaluated the properties of CI (length and coverage) considering single association peaks on one chromosome. However, there might be multiple peaks on one chromosome associated with one trait. Visscher et al., (1996), addressed this problem and proposed to calculate the CI separately, but they never tested if the properties of the CI were similar to the ones for a single QTL. simulated one QTL per plasmode. Thus testing the properties of CI for multiple QTL should be studied. The hypothesis of the proposed study would be that the CI obtained with data partition methods applied independently to multiple QTL will retain the desirable properties shown in Chapter 2 for single QTL. To test the hypothesis, a new plasmode study, simulating two or more QTL peaks in different positions on one chromosome should be performed. The 95% CI of each QTL peak position in a plasmode dataset must be computed using the three proposed methods. Furthermore, coverage should be computed following methods presented in Chapter 2 to determine if the realized coverage of CI is equal to their nominal level. After confirming that the coverage is adequate, the average length of the CI obtained with each method should be compared. According to the GWA results from this study, further work needs to be done for a) resequencing certain regions to find causal variants and improve the annotation, b) validating the proposed candidate genes. Therefore, I propose: 1. Validating the candidate gene CRAT.  This study found a QTL located on SSC1 between 302.9-307.1 Mb. Carnitine O-Acetyltransferase (CRAT) is one of the functional genes annotated in that genomic region. This enzyme is involved in lipid metabolism and it was shown in beef cattle to be   99 differentially expressed in subcutaneous tissue (da Costa et al., 2013). Therefore, I propose a three tier analysis to validate this candidate. First, a differential gene expression analysis can be done with the data of the 176 F2 Duroc x Pietrain animals that is already available in the fat tissue, to compare the expression level of the gene in animals with extreme fat deposition phenotypes. In addition to the in-silico validation, an in vitro experiment could be conducted using adipocytes to assess the potential roll of CRAT in adipocyte growth. Finally, if results of the other two studies are promising, an in vivo study   2. Resequencing the genomic region on SSC6 associated with fat traits. The genomic region between 127.6-140.8 Mb. on SSC6 associated with backfat traits, contains the gene Leptin Receptor Overlapping Transcript (LEPROT), which negatively regulates leptin cell surface exposed receptor (Couturier et al., 2007). Moreover, the Leptin Receptor gene (LEPR) encodes for multiple isoforms of the leptin receptor (Tartaglia, 1997). Causative variants in LEPR have also been associated with carcass measurements such as backfat traits ( Ovilo et al., 2005; Muñoz et al., 2009; Muñoz et al., 2011; Uemoto et al., 2012). Although it has been reported that LEPR maps to SSC6 (Ernst et al., 1996), we were not able to find this gene, because it is located on an unassigned contig in the current pig genome assembly (version 10.2.83) and thus it is not annotated in the current assembly. Therefore, resequencing this genomic region could be beneficial to a) annotate the LEPR gene in that region, b) discover SNP in the gene, c) confirm if those SNP are associated with the phenotypes of interest.    100            LITERATURE CITED   101 LITERATURE CITED   Costa,  a. S. H., Pires, V. M. R., Fontes, C. M. G. a., & Prates, J. a. M. (2013). Expression of genes controlling fat deposition in two genetically diverse beef cattle breeds fed high or low silage diets. BMC Veterinary Research, 9, 118. http://doi.org/10.1186/1746-6148-9-118 Couturier, C., Sarkis, C., Séron, KSilencing of OB-RGRP in mouse hypothalamic arcuate nucleus increases leptin receptor signaling and prevents diet-induced obesity. Proceedings of the National Academy of Sciences of the United States of America, 104(49), 1947619481. http://doi.org/10.1073/pnas.0706671104 Ernst, C. W., Kapke, P. A., Yerle, M., & Rothschild, M. F. (1996). The leptin receptor gene (LEPR) maps to porcine chromosome 6. Mamalian Genome, 8, 226. Goddard, M. E., & Hayes, B. J. (2007). Genomic selection. J.Anim.Breed.Genet, 124, 323330. http://doi.org/10.1080/09064700801959395 Hayes, B. J. (2013). Overview of statistical methods for genome-wide association studies. C. Gondor, B. van der Werf, & B. J. Hayes, editors, Genome-wide association studies and genomic prediction. Human Press, New York, NY. p. 156-157. Hazel, L. N. (1943). The Genetic Basis for Constructing Selection Indexes. Genetics, 28(6), 476490. Jenko, J., Gorjanc, G., Cleveland, M. A., Varshney, R. K., Whitelaw, C. B. A., Woolliams, J. A., & Hickey, J. M. (2015). Potential of promotion of alleles by genome editing to improve quantitative traits in livestock breeding programs. , 47(1), 55. http://doi.org/10.1186/s12711-015-0135-3 Lonergan, S. M., Huff-Lonergan, E., Rowe, L. J., Kuhlers, D. L., & Jungst, S. B. (2001). Selection for lean growth efficiency in Duroc pigs influences pork quality. Journal of Animal Science, 79(8), 20752085. http://doi.org//2001.7982075x Meuwissen, T. H. E., Hayes, B. J., & Goddard, M. E. (2001). Prediction of total genetic value using genome-wide dense marker maps. Genetics, 157(4), 18191829. http://doi.org/11290733 Rodriguez, M. C. (2011). Effects of porcine MC4R and LEPR polymorphisms, gender and Duroc sire line on economic traits in Duroc??Iberian crossbred pigs. Meat Science, 88(1), 169173. http://doi.org/10.1016/j.meatsci.2010.12.018 Muñoz, G., Ovilo, C., Silló, L., Tomás, A., Noguera, J. L., & Rodríguez, M. C. (2009). Single- And joint-population analyses of two experimental pig crosses to confirm quantitative trait loci on Sus scrofa chromosome 6 and leptin receptor effects on fatness and growth traits. Journal of Animal Science, 87(2), 459468. http://doi.org/10.2527/jas.2008-1127   102 (2005). Fine mapping of porcine chromosome 6 QTL and LEPR effects on body composition in multiple generations of an Iberian by Landrace intercross. Genetical Research, 85(1), 5767. http://doi.org/10.1017/S0016672305007330 Tartaglia, L. a. (1997). The leptin receptor. J. Biol. Chem., 272, 60936096. http://doi.org/10.1074/jbc.272.10.6093 Suzuki, K. (2012). Effects of porcine leptin receptor gene polymorphisms on backfat thickness, fat area ratios by image analysis, and serum leptin concentrations in a Duroc purebred population. Animal Science Journal, 83(5), 375385. http://doi.org/10.1111/j.1740-0929.2011.00963.x van Wijk, H. J., Arts, D. J. G., Matthews, J. O., Webster, M., Ducro, B. J., & Knol, E. F. (2005). Genetic parameters for carcass composition and pork quality estimated in a commercial production chain. J. Anim. Sci, 83, 324333. Visscher, P. M., Thompson, R., & Haley, C. S. (1996). Confidence Intervals in QTL Mapping by Bootstrapping. Genetics, 143, 10131020. Weller, J. I., & Ron, M. (2011). Invited review: quantitative trait nucleotide determination in the era of genomic selection. Journal of Dairy Science, 94(3), 10821090. http://doi.org/68/jds.2010-3793