Leveraging pleiotropy to improve genetic risk prediction across diseases.

Abstract

Polygenic scores (PGSs) have shown promise in predicting disease risk, but their predictive accuracy remains limited for many complex diseases. Leveraging the shared genetic architecture among correlated traits may improve prediction performance. We developed a flexible framework for constructing multitrait PGSs by integrating candidate PGSs (N = 2651) derived from 51 genome-wide association studies summary statistics using single-trait, multi-trait analysis of GWAS (MTAG)-all, and MTAG-pairwise approaches. Multitrait PGS models were trained using elastic net regression in the UK Biobank (N = 307,230 individuals) and validated in both an internal set of UK Biobank individuals (N = 39,122) and All of Us (N = 116,394) RESULTS: Multitrait PGSs significantly improved risk prediction for eight diseases, with area under the receiver operating characteristic curve gains ranging from 1.56% to 5.45% compared with optimal single-genome-wide association studies PGSs. Multitrait PGSs further enhanced predictive performance when integrated with nongenetic factors. Significant interactions were identified between multitrait PGS for peripheral artery disease and smoking and waist-to-hip ratio. A clustering analysis uncovered genetically distinct subgroups with meaningful phenotypic variation, including a chronic kidney disease subgroup enriched for diabetes- and obesity-related traits. Our multitrait PGS framework improves disease prediction by capturing cross-trait genetic effects and enables personalized risk assessment through integration with nongenetic exposures, interactions, and subgroup identification.