The most accurate PRS approach varied by the proportion of European ancestry. Populations with greater than 20% African ancestry benefitted significantly from the inclusion of population-specific weights (Figure 2). Intermediate European ancestry individuals benefitted most from using fixed-effects meta-analysis weighting instead of European weights (r = 0.64 versus 0.61, p = 0.02). In contrast, variant weighting from an African ancestry GWAS instead of from European had higher accuracy in low European ancestry (r = 0.65 versus 0.53; p = 0.009) and African-only (r = 0.64 versus 0.45; p = 2.02 × 10−44) populations. Individuals with high European ancestry had similar accuracy with weights from a fixed-effects meta-analysis as from European (r = 0.73 in both; p = 0.79) but decreased performance with the inclusion of weights from an African ancestry GWAS (r = 0.62 versus 0.73; p = 0.01).
