Natural selection acting on complex traits hampers the predictive accuracy of polygenic scores in ancient samples.

The prediction of phenotypes from ancient humans has gained interest due to its potential to investigate the evolution of complex traits. These predictions are commonly performed using polygenic scores computed with DNA information from ancient humans along with genome-wide association study (GWAS) data from present-day humans. However, numerous evolutionary processes could impact these phenotypic predictions. In this work, we investigate how natural selection shapes the temporal dynamics of variants with an effect on the trait and how these changes impact phenotypic predictions for ancient individuals using polygenic scores. We find that stabilizing selection accelerates the loss of large-effect alleles contributing to trait variation. Conversely, directional selection accelerates the loss of small- and large-effect alleles that drive individuals farther away from the optimal phenotypic value. These phenomena result in specific shared genetic variation patterns between ancient and modern populations that hamper the accuracy of polygenic scores to predict phenotypes. Our results assume perfectly estimated effect sizes at the causal loci of complex traits segregating in a GWAS performed in the present and, therefore, provide a putatively loose upper bound on the polygenic score portability to predict traits in the past. Furthermore, we show how natural selection could impact the predictive accuracy of ancient polygenic scores for two widely studied traits: height and body mass index. Our results emphasize the importance of considering decreases on the reliability of polygenic scores to perform phenotypic predictions in ancient individuals due to allele frequency changes driving the loss of alleles via natural selection.