Using PGSs, the proportion of variation in liability to MD explained in European ancestry case-control studies also showed a considerable increase from an R2 of 3.2% in our previous analyses to 5.8% using SBayesR. We also show a significant MD prediction in diverse non-European and admixed ancestries. The SNP-h2 in this study of 8.4% implies that approximately 69% of the additive genetic variance for MD associated with common SNPs across studies can now be accounted for by PGSs. This study provides the first evidence of limited transferability of MD PGS to multiple diverse ancestries and further emphasizes the importance of conducting future GWAS studies across different global populations, especially in Africa, where transferability is poorest. While we did not find evidence for improved prediction based on multi-ancestry rather than European-only PGS, this may be due to the small proportion of participants within each individual ancestry group (23% of individuals of non-European ancestries were divided across 4 major ancestry and admixed groups) relative to the European ancestry group alone.
