30Many diseases and complex traits exhibit population-specific causal effect sizes 31 with trans-ethnic genetic correlations significantly less than 1, limiting trans-ethnic 32 polygenic risk prediction. We developed a new method, S-LDXR, for stratifying 33 squared trans-ethnic genetic correlation across genomic annotations, and applied S-34 LDXR to genome-wide association summary statistics for 30 diseases and complex 35 traits in East Asians (EAS) and Europeans (EUR) (average N EAS =93K, N EUR =274K) 36 with an average trans-ethnic genetic correlation of 0.83 (s.e. 0.01). We determined 37 that squared trans-ethnic genetic correlation was 0.81¢ (s.e. 0.01) smaller than the 38 genome-wide average at SNPs in the top quintile of background selection statistic, 39 implying more population-specific causal effect sizes. Accordingly, causal effect sizes 40 were more population-specific in functionally important regions, including coding, con-41 served, and regulatory regions. In analyses of regions surrounding specifically expressed 42 genes, causal effect sizes were most population-specific for skin and immune genes and 43 least population-specific for brain genes. Our results could potentially be explained 44 by stronger gene-environment interaction at loci impacted by selection, particularly 45 positive selection. 46 2 Trans-ethnic genetic correlations are significantly less than 1 for many diseases and 48 complex traits, 1-6 implying that population-specific causal disease effect sizes contribute to 49 the incomplete portability of genome-wide association study (GWAS) findings and poly-50 genic risk scores to non-European populations. 6-12 However, current methods for estimating 51 genome-wide trans-ethnic genetic correlations assume the same trans-ethnic genetic correla-52 tion for all categories of SNPs, 2,5,13 providing little insight into why causal disease effect sizes 53 are population-specific. Understanding the biological processes contributing to population-54 specific causal disease effect sizes can help inform polygenic risk prediction in non-European 55 populations and alleviate health disparities. 6,14,15 56 Here, we introduce a new method, S-LDXR, for stratifying squared trans-ethnic ge-57 netic correlation across functional categories of SNPs using GWAS summary statistics and 58 population-matched linkage disequilibrium (LD) reference panels (e.g. the 1000 Genomes
59Project 16 ); we stratify the squared trans-ethnic genetic correlation across functional cate-60 gories to robustly handle noisy heritability estimates. We confirm that S-LDXR yields ro-61 bust estimates in extensive simulations. We apply S-LDXR to 30 diseases and complex traits 62 with GWAS summary statistics available in both East Asian (EAS) and European (EUR) 63 populations, leveraging recent large studies in East Asian populations from the CONVERGE 64 consortium and Biobank Japan; 17-19 we analyze a broad set of genomic annotations from the 65 baseline-LD model, 20-22 as well as tissue-specific annotations based o...