Abstract
Purpose :
Glaucoma is a leading cause of irreversible blindness, affecting more than 70 million individuals worldwide. Primary open-angle glaucoma (POAG) is the most common form of glaucoma accounting for approximately 90% of all glaucoma cases. At present, there is no cure for glaucoma and early detection is crucial. We constructed multi-trait polygenic risk scores (PRSs) for glaucoma and tested whether the multi-trait PRSs improved glaucoma prediction.
Methods :
We conducted this study using European participants (n = 435,678) from the UK Biobank data set. We constructed multi-trait PRSs using SNPs derived from the UK Biobank data and previously reported SNPs associated with POAG and its endophenotypes. We examined the associations of the multi-trait PRSs with glaucoma using logistic regression and machine learning methods. To quantify the discriminatory ability of the PRSs on glaucoma, we used the area under the receiver operating characteristic curve (AUC). To avoid overfitting, we used independent training and testing datasets.
Results :
The multi-trait PRSs were significantly associated with glaucoma (P = 8.36e-88), after adjusting for age and sex. Subjects in the top quintile of multi-trait PRSs were 6.52 (P = 2.59e-63) times more likely to have glaucoma, compared with those in the bottom category. The multi-trait PRSs improved the discriminatory power for POAG (AUC increased by 5.5%, P = 1.87e-24) when added to other covariates. Machine learning methods may further improved prediction accuracy.
Conclusions :
We determined that multi-trait PRSs improve the prediction of glaucoma and enhance early detection in genetically susceptible individuals.
This is a 2020 ARVO Annual Meeting abstract.