Localizing components of shared transethnic genetic architecture of complex traits from GWAS summary data

Methods for transethnic genetic analyses rely on assumptions on the transferability of certain components of genetic architecture across ancestry groups (e.g., same causal variants with similar effect sizes in two populations). However, the extent to which these assumptions hold for different complex traits has not been systematically assessed. Here we propose an empirical Bayes framework that models GWAS summary data in two ancestry groups to estimate (1) the prior expected proportions of population-specific/shared causal SNPs and (2) per-SNP posterior probabilities of being causal in one or both populations. Through extensive simulations, we show that our approach yields nearly unbiased estimates of the prior expected proportions of population-specific/shared causal SNPs. We define a statistic to test for enrichment of causal SNPs in a region -- the ratio between the posterior and prior expected numbers of causal SNPs in the region -- and show that it is conservative at different levels of polygenicity and GWAS power. We analyze publicly available GWAS data for 9 complex traits and diseases in individuals of East Asian and European ancestry and find that most high-posterior SNPs are shared and have highly correlated GWAS marginal effects. In ancestry-specific GWAS risk regions, we observe a 2.8x enrichment of SNPs with high posterior probability of being causal in both populations, suggesting that GWAS associations that do not replicate across populations can be explained in part by differences in sample size, causal effect sizes, and/or LD patterns. Finally, we report enrichments of population-specific/shared causal SNPs in 53 tissue-specific functional genomic annotations.