In this study, we used GWAS summary statistics to elucidate shared genetic etiology across psychiatric outcomes by classifying single nucleotide polymorphisms (SNPs) based on their genome-wide association patterns. Specifically, we applied latent profile analysis (LPA) (Gibson 1959; Lazarsfeld and Henry 1968) to p-values of associations between SNPs and three psychiatric phenotypes: symptom counts of alcohol dependence (AD), antisocial personality disorder (ASP), and major depression (MD), to identify clusters of SNPs with homogeneous association patterns. We hypothesized that the pattern of associations with these phenotypes would not be random, but rather, that there would be a subset of SNPs with relatively stronger associations with the two externalizing disorders (AD and ASP) but not with MD, an internalizing disorder. Thus, as a proof of principle, we applied LPA to the association patterns across AD, ASP, and MD, to test whether it would result in identifying clusters of SNPs corresponding to the patterns predicted based on twin findings.
