RT Journal Article SR Electronic T1 Epistatic networks jointly influence phenotypes related to metabolic disease and gene expression in Diversity Outbred mice JF bioRxiv FD Cold Spring Harbor Laboratory SP 098681 DO 10.1101/098681 A1 Anna L. Tyler A1 Bo Ji A1 Daniel M. Gatti A1 Steven C. Munger A1 Gary A. Churchill A1 Karen L. Svenson A1 Gregory W. Carter YR 2017 UL http://biorxiv.org/content/early/2017/04/20/098681.abstract AB Genetic studies of multidimensional phenotypes can potentially link genetic variation, gene expression, and physiological data to create multi-scale models of complex traits. Multi-parent populations provide a resource for developing methods to understand these relationships. In this study, we simultaneously modeled body composition, serum biomarkers, and liver transcript abundances from 474 Diversity Outbred mice. This population contained both sexes and two dietary cohorts. Using weighted gene co-expression network analysis (WGCNA), we summarized transcript data into functional modules which we then used as summary phenotypes representing enriched biological processes. These module phenotypes were jointly analyzed with body composition and serum biomarkers in a combined analysis of pleiotropy and epistasis (CAPE), which inferred networks of epistatic interactions between quantitative trait loci that affect one or more traits. This network frequently mapped interactions between alleles of different ancestries, providing evidence of both genetic synergy and redundancy between haplotypes. Furthermore, a number of loci interacted with sex and diet to yield sex-specific genetic effects. We were also able to identify alleles that potentially protect individuals from the effects of a high-fat diet. Although the epistatic interactions explained small amounts of trait variance, the combination of directional interactions, allelic specificity, and high genomic resolution provided context to generate hypotheses for the roles of specific genes in complex traits. Our approach moves beyond the cataloging of single loci to infer genetic networks that map genetic etiology by simultaneously modeling all phenotypes.