Using Genomic SEM for GWAS of Latent Phenotypes

Genome wide association studies are conventionally conducted using a “massively univariate” approach in which a regression effect is estimated for each individual SNP on a single measured phenotype. A long history of... [ view full abstract ]

Genome wide association studies are conventionally conducted using a “massively univariate” approach in which a regression effect is estimated for each individual SNP on a single measured phenotype. A long history of research in psychometrics and quantitative genetics indicates that many different cognitive, behavioral, and psychiatric phenotypes are genetically correlated. For both theoretical and practical reasons, cross-cutting dimensions of genetic liability across phenotypes (e.g. general intelligence, externalizing psychopathology) may be more appropriate targets for GWAS than the individual phenotypes themselves (e.g. numerical reasoning, vocabulary knowledge, working memory; conduct disorder, alcohol use, sexual risk taking). Here we demonstrate how Genomic SEM can be used to conduct GWAS of latent dimensions of cross-cutting genetic liability. Genomic SEM is freely available to users as a package in R.