Multimodal Prediction of Primary Open-Angle Glaucoma Using Polygenic Risk Scores and Clinical Features in a High-Risk African Ancestry Cohort
In the authors' words
Primary open-angle glaucoma (POAG) disproportionately affects individuals of African ancestry, for whom risk prediction tools remain limited. We evaluated four polygenic risk scores (PGS), two genome-wide and two curated, weighted by African ancestry effect sizes. Used alone, the scores showed modest discrimination. Age and sex were strongly predictive; adding PGS616 reached a cross-validated AUC of 0.713 in the training cohort (N = 271), comparable to the demographic model. In the Penn Medicine BioBank (N = 9,084) the model reached AUC = 0.727, similar to 0.729 for age and sex alone, with comparable case detection at fixed specificity. In a suspect cohort (N = 1,013), predicted risk tracked cup-to-disc ratio and nerve fiber layer thinning, an association explained by age and sex. These results establish a demographic benchmark for evaluating ancestry-matched PGS in POAG. HIGHLIGHTSO_LIAge and sex alone predict POAG well in a high-risk African ancestry cohort C_LIO_LIFour ancestry-matched PGS show modest standalone discrimination in this cohort C_LIO_LIAdding PGS616 to age and sex gives comparable AUC in both cohorts (primary MLP) C_LIO_LIPredicted risk tracks optic nerve damage in suspects, driven by age and sex C_LI
Appeared: Tuesday, September 22. medRxiv. Preprint, not yet peer-reviewed.