Endophenotype-Informed Polygenic Risk Stratification for Primary Open-Angle Glaucoma in individuals of African Ancestry

Primary open-angle glaucoma (POAG) is a leading cause of irreversible blindness, yet up to half of cases remain undiagnosed, limiting opportunities for early intervention. Genetic risk stratification may improve early detection, particularly in people of African ancestry, who experience a disproportionate burden of POAG but remain underrepresented in genetic studies. Prior glaucoma polygenic risk score (PGS) approaches have incorporated POAG and other related ocular traits, including intraocular pressure (IOP) and cup-to-disc ratio (CDR) to improve prediction. However, most evaluations have emphasized aggregate discrimination, while less is known about whether endophenotype-specific genetic information can improve clinical interpretability or identify distinct individuals at the extremes of risk. Here, we developed endophenotype-informed PGS models that preserve genetic information related to IOP and CDR as distinct components of risk. These scores were evaluated in an independent testing cohort from the Primary Open-Angle African Ancestry Glaucoma Genetics (POAAGG) study (n=271), validated in the Penn Medicine Biobank (PMBB) cohort (n=1,662), and assessed for clinical relevance in a longitudinal cohort of POAAGG glaucoma suspects. We also performed manual chart reviews of PMBB participants in the highest and lowest extremes (5th percentiles) of PGS distribution to evaluate whether endophenotype-informed scores identified clinically meaningful patterns of risk. Endophenotype-informed PGS provide pathway-level attribution not available from a single aggregate POAG PGS score. Although models incorporating measured IOP and CDR as covariates achieved similar overall discrimination, they did not prioritize the same individuals at the extremes of predicted risk. In the longitudinal suspect cohort, the baseline IOP-informed score stratified individuals who later converted to POAG from those who remained stable. Manual chart review further supported the clinical interpretability: the CDR-informed score identified structural susceptibility to glaucomatous damage beyond EHR-based classification, while the IOP-informed score captured genetic susceptibility to pressure-related disease. Together, these findings show that endophenotype informed polygenic modeling can move glaucoma genetic risk prediction beyond aggregate discrimination toward a more clinically interpretable framework for risk stratification. In African ancestry populations, this approach may help identify biologically meaningful heterogeneity, support conversion-risk prediction, and prioritize individuals who could benefit from earlier surveillance.