Summary statistics from iPSYCH2015 and the following external samples were included in a fixed-effects variance-weighted meta-analysis in METAL28: 1) 230,118 broadly defined depression cases and 545,339 controls from Howard et al. 201914, including self-reported depression of 23andMe23 and the broadly defined depression phenotype of UKB, 35,077 narrowly defined depression cases and 95,406 controls from Wray et al. 201812, excluding 23andMe23; 2) depression phenotypes from the Million Veteran Program (MVP)24 based on ICD codes derived from electronic health records (83,810 cases and 166,405 controls); 3) 28,098 cases with ICD-10 F32 and/or F33 diagnoses and 228,817 controls from FinnGen22. For comparison, we also conducted a GWAS and several downstream analyses using a narrower definition of depression that excluded self-reported depression of 23andMe12,23 and the broadly defined depression phenotype of UKB14,21 (Supplementary Figure S14, Table S13). The results of the downstream analyses were very similar to the primary analyses without noteworthy differences, except for the DEP-PRS analysis of cognitive performance in the PNC cohort, which is mentioned in the results section.
