between our discovery and replication GWASs as an overall metric of replicability3,17. Among the 11,529 out of 11,847 SNPs that had a MAF greater than 1% in the EBB, we found a correlation of marginal SNP effects of ρb = 0.93 (jackknife standard error; s.e. 0.01) and a correlation of conditional SNP effects using the same LD reference panel of ρb = 0.80 (s.e. 0.03; Supplementary Fig. 14). Although we had limited power to replicate associations with 238 GWS variants that are rare in the EBB (MAF < 1%), we found, consistent with expectations (Methods and Extended Data Fig. 3b), that 60% of them had a marginal SNP effect that was sign-consistent with that from our discovery GWAS (Fisher's exact test; P = 0.001). The proportion of sign-consistent SNP effects was greater than 75% (Fisher's exact test; P < 10−50) for variants with a MAF greater than 1%—also consistent with expectations (Extended Data Fig. 3b). Altogether, our analyses demonstrate the robustness of our findings and show their replicability in an independent sample.
