Integrating Glaucoma Endophenotypes Improves Polygenic Risk Prediction for Primary Open-Angle Glaucoma Across Ancestries

Abstract

Primary open-angle glaucoma (POAG) is a leading cause of irreversible blindness, and early detection is crucial. Existing polygenic risk scores (PRSs) rely on single-trait models that lack the complementary genetic information captured by glaucoma endophenotypes. We developed a multi-trait polygenic probability risk score (PPRS) framework integrating trait-specific PRSs to improve prediction across ancestries. We evaluated the PPRS in the UK Biobank (n=324,713, European ancestry) and the Mexican American Glaucoma Genetic Study (MAGGS, n=4,549, Latino ancestry). Using SBayesRC with > 7 million variants and 96 functional annotations, we constructed PRSs for POAG, intraocular pressure (IOP), vertical cup-to-disc ratio (VCDR), and retinal nerve fiber layer thickness. Our PPRS combines these trait-specific PRSs via penalized logistic regression, preserving each traits contribution. The PPRS achieved AUC values of 0.814 in Europeans and 0.802 in Latinos. Individuals in the highest PPRS decile had 74.4-fold (Europeans) and 49.3-fold (Latinos) higher odds of POAG than those in the lowest decile, with the top quin-tile capturing 65.7% and 62.2% of cases, respectively. Component trait contributions differed by ancestry: IOP PRS was strongest in Europeans (OR=1.63, P = 5.37 x 10-89), whereas VCDR PRS dominated in Latinos (OR=1.64, P = 2.04 x 10-11). Our multi-trait PPRS framework improves POAG prediction across ancestries and reveals that distinct endophenotypes drive genetic risk in different populations, supporting ancestry-informed risk stratification for targeted screening and earlier intervention. By decomposing polygenic risk into trait-specific components, this framework enables ancestry-informed glaucoma screening strategies that could facilitate earlier detection and personalized clinical management.