Polygenic Scores (PSs) are computed by summing the contribution of many associated alleles across the genome1,2. These contributions are weighted by allele effect sizes. Such effect sizes are extrapolated from Genome Wide Association Studies (GWAS)3,4 carried out in a specific population, or across multi-ethnic sample sets5–7. Most of the times, associated SNPs are thus merely correlated with a phenotype and are not really causal3. Furthermore, as environment interactions, Linkage Disequilibrium patterns, allele frequencies and rare polymorphisms are often population specific, effect sizes might be in part population dependent3,8,9. This may lead PS to exhibit a directional bias and a lower predictivity in individuals from populations not closely related to the one where the GWAS study was performed10–14 or even from the same population as it was shown for UK and Finnish cohorts15–17. This is particularly problematic with recently admixed individuals, where the various ancestries composing a given genome may be closely or distantly related to the population used to infer the adopted genetic effect sizes18. Admixed individuals are indeed expected to constitute a considerable portion of contemporary and future societies19. Therefore, it is crucial to include them within the promising vault of the emerging predictive and personalized healthcare.
