Using Genomic SEM for Multivariate GWAS Discovery

A powerful application of Genomic Structural Equation Modeling (Genomic SEM) is to specify a model in which the SNP effects occur at the level of a latent genetic factor defined by several phenotypes. This allows researchers... [ view full abstract ]

A powerful application of Genomic Structural Equation Modeling (Genomic SEM) is to specify a model in which the SNP effects occur at the level of a latent genetic factor defined by several phenotypes. This allows researchers to identify variants with effects on general dimensions of cross-trait liability (i.e., multivariate GWAS), boost power for discovery, and calculate more predictive polygenic scores. When certain SNPs only influence a subset of genetically correlated traits, a key assumption of the model is violated. With this in mind, we develop a test of heterogeneity that can be used to evaluate the extent to which the effect of a given SNP operates through a common pathway(s) model. SNPs with high heterogeneity estimates can be flagged as likely to confer disproportionate or specific liability toward individual traits or disorders, can be removed when constructing polygenic risk scores, or studied specifically to understand the nature of heterogeneity. These heterogeneity estimates act as safeguards against false inference when considering a locus specific to one trait in its effect on a set of correlated traits. As an example of this approach, we conduct a joint analysis of GWAS summary statistics from five genetically correlated psychiatric case-control traits: schizophrenia, bipolar disorder, major depressive disorder (MDD), post-traumatic stress disorder (PTSD), and anxiety. We model genetic covariances among the traits using a general factor of psychopathology (p), for which we identify 27 independent SNPs not previously identified in the univariate GWASs, 5 of which can be validated based on previous outside GWASs. Polygenic scores derived using this p-factor consistently outperform polygenic scores derived from GWASs of the individual traits in out-of-sample prediction of psychiatric symptoms. As a tool for discovery and validation, Genomic SEM is both flexible and allows for continuous innovations in how multivariate genetic architecture is modeled.