We acquired GWAS summary statistics for each of 21 shared traits between EUR and EAS for which there was at least one significant IMPACT association in each population. Then, we restricted to SNPs shared between EUR and EAS GWAS summary statistics. Next, we performed stringent iterative LD clumping with PLINK (v.1.90b3)63 using EUR summary statistics (selecting the most significant SNP, then removing all SNPs in LD with r2 > 0.1 within 1 Mb, then selecting the next most significant SNP and so on). We selected our initial set of SNPs under three scenarios: (1) using no functional inference, (2) using the top 5% of SNPs according to the trait’s lead EUR IMPACT annotation and (3) using the bottom 95% of SNPs according to the trait’s lead EUR IMPACT annotation (mutually exclusive with scenario 2). With our set of independent SNPs for each trait and under each of three scenarios, we compute a Pearson correlation between the estimated effect sizes, while further stratifying loci on 17 EUR P values (1, 0.3, 0.1, 0.03, 0.01, 3 × 10−3, 1 × 10−3, 3
