Multimodal Prediction of Primary Open-Angle Glaucoma Using Polygenic Risk Scores and Clinical Features in a High-Risk...

Area of Science:

• Ophthalmology
• Genetics
• Computational Biology

Background:

• Primary open-angle glaucoma (POAG) disproportionately impacts individuals of African ancestry.
• Current early detection tools for POAG are limited, especially in at-risk populations.

Purpose of the Study:

• To develop and validate machine learning models for early POAG risk stratification in African ancestry populations.
• To assess the utility of ancestry-matched polygenic risk scores (PRS) integrated with clinical features.

Main Methods:

• Utilized the largest African ancestry POAG cohort (N=7,352) for model development.
• Integrated clinical features with four ancestry-matched PRS (two genome-wide, two curated).
• Trained models on a clean cohort (N=196) and tested in 1,013 suspects.

Main Results:

• Machine learning models incorporating PRS showed improved prediction accuracy for POAG.
• Curated PRS (MEGA PRS, MTAG PRS) demonstrated particular effectiveness.
• MEGA PRS predictions correlated with key clinical indicators: intraocular pressure (IOP), cup-to-disc ratio (CDR), and retinal nerve fiber layer (RNFL) thickness.
• Integration of PRS with inter-eye asymmetry features further enhanced predictive performance.

Conclusions:

• Curated polygenic risk scores, such as MEGA PRS and MTAG PRS, are valuable, interpretable, and equitable tools for early glaucoma risk stratification.
• These findings support the use of PRS in enhancing early POAG detection in underrepresented populations.