Ancestry outliers (10%) were removed to avoid confounding from population substructure and admixture. To assess generalizability of our main results, we conducted depression-GWAS of the ancestry-outliers using the Ricopili and adjusting for the first 10 principal components (Supplementary Figure S24A and S24B). This GWAS showed a high genetic correlation (rg=0.95, SE=0.24, P=1x10–4) with our primary depression meta-analysis as shown in Supplementary Table S23. Sign-tests of overall replication of the direction of effects for various GWAS P-value thresholds using our primary meta-analysis as discovery sample and the iPSYCH2015 ancestry outliers as replication sample, showed an overall replication ratio of 0.59-0.64 (Supplementary Table S24). This suggests that many more findings would likely replicate, provided a much larger and more powerful sample, in other non-European ancestries. Previous GWAS of depression using trans-ancestral meta-analysis24 generally supports the transferability of GWAS results across ancestries.
