Overall plotting (e.g., regional association) and annotation of individual loci was conducted in FUMA48. For gene and gene-set based analysis, MAGMA, as implemented in FUMA, was used. FUMA was utilized to conduct gene-set analyses that examined whether genes were enriched in curated classification systems, by molecular function, biological process or other criteria. Gene sets were defined for 4,728 curated gene sets (including canonical pathways) and 6,166 GO terms. Differential expression of prioritized genes was conducted using the GENE2FUNC option in FUMA, which examines whether genes of interest from the GWAS are overrepresented in differentially expressed gene sets in 53 specific tissue types from The Genotype-Tissue Expression (GTEx) database49. Although this database is comprised of primarily EA individuals, it is one of few publicly available databases available, and therefore was utilized for the AA results as well. To further prioritize possible causal genes, we used S-PrediXcan50 to impute genetically-regulated gene expression in twelve brain tissues and whole blood. The prediction models were trained on reference transcription data from GTEx (brain) and the Depression Genes and Network (DGN) (whole blood) (all available
