The impact of diverse ancestry on polygenic risk score-based prediction models for psychiatric and neurodevelopmental disorders

Psychiatry Res. 2026 Jul 1;364:117322. doi: 10.1016/j.psychres.2026.117322. Online ahead of print.

ABSTRACT

Precision psychiatry aims to include genetic risk assessed via polygenic risk scores (PRSs) in risk prediction models. A major challenge to health equality is that the predictive accuracy of PRSs decreases when the training and target populations are of different ancestries. However, no prior study has systematically investigated the degree of reduction or loss of specificity for multiple psychiatric disorders. We used two independent waves of a psychiatric case-cohort sample (iPSYCH) to investigate this. We performed GWASs in one wave and assessed the association of PRSs trained on these GWASs with seven psychiatric and neurodevelopmental phenotypes in the other wave. In contrast to previous studies, we included a core ancestry group as well as individuals of diverse ancestry, rather than homogeneous groups of different ancestries. The PRSs were significant predictors for all disorders when both the training and target datasets were of European ancestry, but, for several disorders, we observed a sharp decrease in PRS accuracy when the target group comprised individuals of diverse ancestry relative to the core group, with the a reduction in R² of ∼60% for overall psychiatric diagnosis, and 40-50% for ASD and ADHD. PRSs for psychiatric disorders show population specificity, with large reductions in accuracy when predicting in individuals of diverse ancestry. However, there may be some benefit in including diverse ancestries also in the training dataset for the purpose of improving prediction accuracy. Our results emphasize the importance of the availability of genetic datasets for non-European populations in the context of mental health.

PMID:42418907 | DOI:10.1016/j.psychres.2026.117322