Three open questions in polygenic score portability

TL;DR

Polygenic scores lose accuracy when applied to people genetically different from the original study group, and this issue is influenced by factors like trait type and social context.

Contribution

The study identifies three key gaps in understanding polygenic score portability across diverse populations.

Findings

Individual prediction accuracy is weakly linked to genetic distance but strongly influenced by socioeconomic factors.

Prediction accuracy for immunity-related traits drops sharply at moderate genetic distances, possibly due to fast-evolving genetic variants.

Measures of predictive performance like precision and recall can show conflicting trends with genetic distance.

Abstract

The broad adoption of polygenic scores (PGS) is hindered by their limited portability to people that differ—in genetic ancestry or other characteristics—from the GWAS samples used to construct them. Here, we measure PGS prediction accuracy as a continuous function of individuals’ genome-wide genetic dissimilarity to the GWAS sample (genetic distance). Our results highlight three gaps in our understanding of PGS portability. First, variation in individual-level prediction accuracy is only weakly predicted by genetic distance. In fact, it is explained comparably well by socioeconomic measures. Second, trends of portability vary across traits. For several immunity-related traits, prediction accuracy drops near zero even at intermediate genetic distances—potentially reflecting fast evolutionary turnover of genetic variants associated with immunity. Third, even qualitative trends of portability can depend on how we measure predictive performance. For instance, for type 2 diabetes, precision remains roughly constant, while recall surprisingly increases with genetic distance. Together, our results show that portability cannot be understood through global ancestry groupings alone. Other, understudied factors influence portability, including the specifics of trait evolution, genetic architecture, social context, and the construction of the PGS. Addressing these gaps can aid in the development of PGS and inform more equitable applications.

Genetic predictors of health outcomes often drop in accuracy when applied to people dissimilar to participants of large genetic studies. Here, the authors investigate the root causes and highlight open questions underlying this problem.

Figures