Current methods for improving PRS accuracy in diverse populations have prioritized the inclusion of variants from European GWASs, as these have higher statistical power to identify trait-associated loci. For example, one such approach uses a two-component linear mixed model to allow for the incorporation of ethnic-specific weights.6 Another method creates ancestry-specific partial PRSs for each individual based on the local ancestry of variants selected from a European GWAS.7 This approach was found to improve trait predictability, compared to a traditional PRS with population-specific or European weights, in East Asians for BMI but not height.7 In contrast, implementing this local-ancestry method7 in our simulation, we found that PRS accuracy was higher with African or fixed-effects meta-analysis weighting than with local ancestry in admixed African ancestry populations. Our results suggest that true equality in performance will be difficult to obtain solely based on European-identified variants even with local ancestry-adjusted weights. Although local ancestry weighting may have greater benefits when complete sharing across populations is not assumed, we show that in the absence of population-specific factors, the optimal PRS approach involves using variants identified in a large African population and population-specific weighting.
