Exploiting inherent interdependencies among traits for genetic association analysis
Authors: Haoran Cai; David Des Marais
Affiliation: Massachusetts Institute of Technology, Cambridge MA
Keywords: Complex traits
Quantitative trait locus (QTL) mapping and genomic prediction are pivotal for quantitative genetics and crop improvement. Published studies are almost always univariate, considering each trait separately. Problems remain in how to properly exploit the interdependency among traits in QTL mappings. The aim underlying such joint modeling is to increase the statistical power. However, simply using phenotypic or genetic correlation to improve power for QTL analysis may incur problems. Because the correlation between traits does not imply information in a single SNP level. It is the combination of QTLs that shape the interdependency among traits: a strong correlation between traits need not result in a strong pleiotropy of individual QTLs. Here, by leveraging the environmental correlation, we propose a multivariate framework to detect causal QTLs and systematically assess the pleiotropic structure of genotype-phenotype map. The key assumption underlying our approach is that the overlapped covariance structure of environmental and genetic variation provides the inherent interdependencies and modular nature of trait combination. We will demonstrate how our approach provides interpretable results and causal understandings of genetic architecture.