To examine potential polygenic heterogeneity across depression sub-phenotypes, we investigated how PRS trained on the different phenotypes described above were distributed across depression sub-phenotypes in iPSYCH using a multivariate PRS approach both in a sex-stratified and in an unstratified analysis. The method is described in detail in Grove et al.15, and is a regression of multiple outcome variables, and in principle a linear regression for each PRS on the depression sub-phenotypes allowing for comparisons on the average PRS across sub-phenotypes for PRS from a number of phenotypes adjusting for necessary covariates. Since the variance-covariance matrix is fitted jointly and accounting for the covariance the inherent correlation between scores is adjusted for when estimating variance and in the hypothesis testing. We performed four mvPRS analyses testing for differences among depression-subtypes; depression cases who have developed (1) anxiety, (2) bipolar disorder, (3) schizophrenia or (4) SUD. Each of the four multiple regression analyses tests for individual and distinct hypotheses, one for each PRS, and the reported P-values are not corrected for multiple testing. Bipolar disorder cases were excluded from the analyses not involving bipolar disorder.
