Generalization and Dilution of Association Results from European GWAS in Populations of Non-European Ancestry: The PAGE Study

Abstract

A multi-ethnic study demonstrates that the extrapolation of genetic disease risk models from European populations to other ethnicities is compromised more strongly by genetic structure than by environmental or global genetic background in differential genetic risk associations across ethnicities.

Abstract

The vast majority of genome-wide association study (GWAS) findings reported to date are from populations with European Ancestry (EA), and it is not yet clear how broadly the genetic associations described will generalize to populations of diverse ancestry. The Population Architecture Using Genomics and Epidemiology (PAGE) study is a consortium of multi-ancestry, population-based studies formed with the objective of refining our understanding of the genetic architecture of common traits emerging from GWAS. In the present analysis of five common diseases and traits, including body mass index, type 2 diabetes, and lipid levels, we compare direction and magnitude of effects for GWAS-identified variants in multiple non-EA populations against EA findings. We demonstrate that, in all populations analyzed, a significant majority of GWAS-identified variants have allelic associations in the same direction as in EA, with none showing a statistically significant effect in the opposite direction, after adjustment for multiple testing. However, 25% of tagSNPs identified in EA GWAS have significantly different effect sizes in at least one non-EA population, and these differential effects were most frequent in African Americans where all differential effects were diluted toward the null. We demonstrate that differential LD between tagSNPs and functional variants within populations contributes significantly to dilute effect sizes in this population. Although most variants identified from GWAS in EA populations generalize to all non-EA populations assessed, genetic models derived from GWAS findings in EA may generate spurious results in non-EA populations due to differential effect sizes. Regardless of the origin of the differential effects, caution should be exercised in applying any genetic risk prediction model based on tagSNPs outside of the ancestry group in which it was derived. Models based directly on functional variation may generalize more robustly, but the identification of functional variants remains challenging.

Author Summary

The number of known associations between human diseases and common genetic variants has grown dramatically in the past decade, most being identified in large-scale genetic studies of people of Western European origin. But because the frequencies of genetic variants can differ substantially between continental populations, it's important to assess how well these associations can be extended to populations with different continental ancestry. Are the correlations between genetic variants, disease endpoints, and risk factors consistent enough for genetic risk models to be reliably applied across different ancestries? Here we describe a systematic analysis of disease outcome and risk-factor–associated variants (tagSNPs) identified in European populations, in which we test whether the effect size of a tagSNP is consistent across six populations with significant non-European ancestry. We demonstrate that although nearly all such tagSNPs have effects in the same direction across all ancestries (i.e., variants associated with higher risk in Europeans will also be associated with higher risk in other populations), roughly a quarter of the variants tested have significantly different magnitude of effect (usually lower) in at least one non-European population. We therefore advise caution in the use of tagSNP-based genetic disease risk models in populations that have a different genetic ancestry from the population in which original associations were first made. We then show that this differential strength of association can be attributed to population-dependent variations in the correlation between tagSNPs and the variant that actually determines risk—the so-called functional variant. Risk models based on functional variants are therefore likely to be more robust than tagSNP-based models.