We explored the impact of negative selection through additional simulations by considering different relationships between effect sizes and allele frequencies of causal variants. We found that predictions from our method can underestimate the RA by up to −8.5% in populations undergoing mild negative selection (Fig. 2). Therefore, our predictions can be interpreted as upper bounds for how much RA can be reduced simply because of MAF and LD differences between ancestries. More generally, since PGSs in the current study are mainly focused on common variants, the impact of low frequency and rare variants which may better inform selection remains to be investigated.
