strikingly similar, with a regression coefficient that is consistent with identity (slope = 0.98, s.e.m. = 0.04). For example, we observed a strong Pearson correlation of τ* between EUR and EAS for asthma (r=0.98), RA (r=0.87), MCV (r=0.96), PrCa (r=0.90) and height (r = 0.96). Cross-ancestry functional concordance is not specific to IMPACT annotations as we observed a similar relationship among cell-type-specific histone marks using the same strategy (Supplementary Fig. 11)24. However, we did not observe cross-ancestry concordance for 513 cell-type-specifically expressed gene sets (SEG)24,52, possibly due to a lack of significant associations shared between populations. Furthermore, we found that none of our τ* estimates show evidence of population heterogeneity (all two-tailed difference of means FDR > 0.56). Overall, our results suggest that regulatory variants in EUR and EAS populations are similarly enriched within the same classes of regulatory elements defined by IMPACT. This does not exclude the possibility of population-specific variants or causal effect sizes, as evidenced by 13 traits with trans-ancestry genetic correlation significantly less than 1 (P< 0.05/29 tested traits). Rather, these results suggest that causal biology including disease-driving cell types and associated regulatory elements is largely shared between these populations.
