Cross-ancestry genome-wide association studies identified heterogeneous loci associated with differences of allele frequency and regulome tagging between participants of European descent and other ancestry groups from the UK Biobank

Abstract

To investigate cross-ancestry genetics of complex traits, we conducted a phenome-wide analysis of loci with heterogeneous effects across African, Admixed-American, Central/South Asian, East Asian, European and Middle Eastern participants of the UK Biobank (N = 441 331). Testing 843 phenotypes, we identified 82 independent genomic regions mapping variants showing genome-wide significant (GWS) associations (P < 5 × 10−8) in the trans-ancestry meta-analysis and GWS heterogeneity among the ancestry-specific effects. These included (i) loci with GWS association in one ancestry and concordant but heterogeneous effects among the other ancestries and (ii) loci with a GWS association in one ancestry group and an experiment-wide significant discordant effect (P < 6.1 × 10−4) in at least another ancestry. Since the trans-ancestry GWS associations were mostly driven by the European ancestry sample size, we investigated the differences of the allele frequency (ΔAF) and linkage disequilibrium regulome tagging (ΔLD) between European populations and the other ancestries. Within loci with concordant effects, the degree of heterogeneity was associated with European–Middle Eastern ΔAF (P = 9.04 × 10−6) and ΔLD of European populations with respect to African, Admixed-American and Central/South Asian groups (P = 8.21 × 10−4, P = 7.17 × 10−4 and P = 2.16 × 10−3, respectively). Within loci with discordant effects, ΔAF and ΔLD of European populations with respect to African and Central/South Asian ancestries were associated with the degree of heterogeneity (ΔAF: P = 7.69 × 10−3 and P = 5.31 × 10−3, ΔLD: P = 0.016 and P = 2.65 × 10−4, respectively). Considering the traits associated with cross-ancestry heterogeneous loci, we observed enrichments for blood biomarkers (P = 5.7 × 10−35) and physical appearance (P = 1.38 × 10−4). This suggests that these specific phenotypic classes may present considerable cross-ancestry heterogeneity owing to large allele frequency and LD variation among worldwide populations.

Modeling the longitudinal changes of ancestry diversity in the Million Veteran Program

“…The consistent ancestry assignment performed in these cohorts therefore can permit meta-analyses across two of the largest genetic data repositories in the world in a harmonized fashion. A harmonized definition of ancestry across datasets reduces heterogeneity across the meta-analyzed datasets, increasing the statistical power of the gene discovery analysis [ 19 , 20 ]. However, population stratification among the recently admixed American groups (e.g., African Americans and Latin Americans) still requires careful consideration for within-population adjustment of ancestry diversity [ 4 – 6 ].…”

Sex differences in the polygenic architecture of hearing problems in adults

“…Due to the limited non-EUR sample, we replicated only few single-variant associations and none of them were present across multiple ancestries. While the low statistical power likely played a key role in the lack of multiple-ancestry single-variant replications, differences in allele frequency and LD structure may also have contributed [ 56 ]. Conversely, we observed highly statistically significant associations of HP PRS derived from the EUR discovery GWAS meta-analysis and tested on non-EUR participants.…”