How genetic disease risks can be misestimated across global populations

Abstract

Background: Accurate assessment of health disparities requires unbiased knowledge of genetic risks in different populations. Unfortunately, most genome-wide association studies use genotyping arrays and European samples. Here, we integrate whole genome sequence data from global populations, results from thousands of GWAS, and extensive computer simulations to identify how genetic disease risks can be misestimated.Results: In contrast to null expectations, we find that risk allele frequencies at known disease loci are significantly different for African populations compared to other continents.Strikingly, ancestral risk alleles are found at 9.51% higher frequency in Africa and derived risk alleles are found at 5.40% lower frequency in Africa. By simulating GWAS with different study populations, we find that non-African cohorts yield disease associations that have biased allele frequencies and that African cohorts yield disease associations that are relatively free of bias. We also find empirical evidence that genotyping arrays and SNP ascertainment bias contribute to continental differences in risk allele frequencies. Because of these causes, polygenic risk scores can be grossly misestimated for individuals of African descent. Importantly, continental differences in risk allele frequencies are only moderately reduced if GWAS use whole genome sequences and hundreds of thousands of cases and controls. Finally, comparisons between uncorrected and corrected genetic risk scores reveal the benefits of considering whether risk alleles are ancestral or derived.Conclusions: Our results imply that caution must be taken when extrapolating GWAS results from one population to predict disease risks in another population.

Population differentiation of polygenic score predictions under stabilizing selection

“…Therefore, an allele that is rare in the GWAS sample but common elsewhere will not be discovered, leading to a greater reduction in the variance accounted for, or prediction accuracy, in non-represented populations. (Liu et al, 2015;Martin et al, 2017a,b;Curtis, 2018;Kim et al, 2018;Bentley et al, 2019;Wojcik et al, 2019;Wang et al, 2020;Conti et al, 2021;Cavazos and Witte, 2021). The impact of genetic differentiation on prediction accuracy in non-represented populations can be large, as it has been found that many variants contributing to trait variation in European GWAS samples were not at a high enough frequency to be detected in other populations, suggesting different alleles contribute greatly to the trait variance in each population (Liu et al, 2015;Durvasula and Lohmueller, 2021).…”

“…Even with perfectly estimated effects at the causal loci with significant trait associations, the prediction accuracy of polygenic scores will be limited because GWAS only identify loci with common alleles that contribute enough variance to exceed some significance threshold determined by the sample size. Therefore, an allele that is rare in the GWAS sample but common elsewhere will not be discovered, leading to a greater reduction in the variance accounted for, or prediction accuracy, in unrepresented populations (Martin et al ., 2017a,b; Curtis, 2018; Kim et al ., 2018; Bentley et al ., 2019; Wojcik et al ., 2019; Wang et al ., 2020; Conti et al ., 2021). Indeed, many variants contributing to trait variation in European GWAS samples are not at a high enough frequency to be detected in other populations, suggesting different sets of polymorphisms contribute to the trait variance in different populations (Liu et al ., 2015; Durvasula and Lohmueller, 2021).…”

Population differentiation of polygenic score predictions under stabilizing selection

“…Therefore, an allele that is rare in the GWAS sample but common elsewhere will not be discovered. This would lead to a greater reduction in the phenotypic variance accounted for, or prediction accuracy, in populations not represented in the GWAS sample (hereafter 'unrepresented populations'; [30,40,44,[48][49][50][51][52]). Indeed, many variants contributing to trait variation in European GWAS samples are not at a high enough frequency to be detected in other populations, suggesting that different sets of polymorphisms contribute to the trait variance in different populations [53,54].…”