Multimodal Prediction of Primary Open-Angle Glaucoma Using Genetic Risk Scores and Clinical Features in a High-Risk African Ancestry Cohort

Primary open angle glaucoma (POAG) disproportionately affects individuals of African ancestry, yet early detection tools remain limited. Using the largest African ancestry POAG cohort (N=7,352), we developed an integrative machine learning framework combining curated genetic risk scores (GRS 1 to 4), polygenic risk scores (PRS), and demographic features. A clean, biologically extreme training cohort (N=196) enabled robust model development with Random Forest, Support Vector Machine, and Multi Layer Perceptron classifiers, followed by validation in 1,013 unlabeled glaucoma suspects. Curated GRS, particularly GRS 2 and GRS 4, consistently outperformed PRS and significantly boosted predictive accuracy (AUC up to 0.95 when combined with base features). Risk predictions correlated with intraocular pressure, cup-to-disc ratio, and retinal nerve fiber layer thickness in suspects, confirming biological relevance. These results establish curated, ancestry-matched GRS as interpretable and clinically translatable predictors, paving the way for equitable, genetics driven precision screening strategies in high risk populations.