While the best single PRS for admixed individuals with at least 20% African ancestry selected variants based on a GWAS in an African ancestry population with weights from a fixed-effects meta-analysis, a linear combination of the European and African ancestry-derived PRSs had higher accuracy; this was particularly true at decreased African ancestry cohort sizes. We saw considerable improvements with the combined PRS over using a European-derived (European selected variants and weights) PRS, especially for low European ancestry (CEU < 20%), where even with 10-fold fewer African samples there was a 27.4% increase in PRS accuracy compared to the European-derived risk score and a 12.3% increase compared to a PRS with African ancestry selected variants and weights from a fixed-effects meta-analysis (Figure 3).
