Before introducing differential ancestry effects in our system we tested whether a PS computed on a partial genome (pPS) can be used as proxy for the total PS and as predictor of the expressed phenotype. We relied on genomes for which phenotypes (T2D, breast cancer, BMI, height) and total PS were both available: the Estonian Biobank31 (EstBB). We mimicked a history of non-European admixture in EstBB samples by applying onto them the local ancestry patterns resulting from the analysis in the previous section. Importantly, even if it has been shown that PS performance can vary even among European populations15,16, we do not expect macroscopic inefficiencies adopting available GWAS results in Estonians12,18. We therefore investigated whether the pPS could be used to obtain a significant improvement in the phenotype prediction, compared to the baseline linear model calculated with no genetic information. For all traits, we show that pPS calculated with even a small portion of available genome are capable of significantly improving the phenotype prediction accuracy, to an extent that depends on the SNPs included in that portion and its size
