Overall, we constructed twelve PRSs with variants selected from GWASs in European or African ancestry populations or a fixed-effects meta-analysis of both (three scenarios) and weights from the same approaches plus an additional local ancestry-specific weighting method (four scenarios) (Figure 2). For Europeans, the highest PRS accuracy was achieved with European selected variants and weights (r = 0.77; 95% CI = [0.76, 0.77]); however, a similar accuracy was observed for weights from a fixed-effects meta-analysis (r = 0.76; p = 0.53). For Africans, the highest PRS accuracy was with African selected variants and weights from a fixed-effects meta-analysis (r = 0.75; 95% CI = [0.74, 0.75]); similar performance was observed with African variants and weights (r = 0.74; p = 0.28). For admixed individuals, the highest-performing PRS depended on the population ancestry percentage. In individuals with high European ancestry (>80%), the best PRS was with European selected variants and fixed-effects meta-analysis or European weights (r = 0.73; 95% CI = [0.72, 0.74]). For individuals with intermediate (20%–80%) or low (<20%) European ancestry, the most accurate PRS was from using African
