Fig. 3
From: Multimodal deep learning approaches for improving polygenic risk scores with imaging data
Prediction accuracy of single predictors and multimodal approaches for glaucoma using XGBoost, with AUC values displayed in ascending order. (a) Bar plot of AUC values for single predictors, including PRS, IDS, IDPs, IGSs, and clinical covariates, in individuals with OCT images (N=55,469). (b) Bar plot of AUC values for multimodal approaches, including PRS + IDS, PRS + IGSs, PRS + IGSs + clinical covariates, and PRS + IDS + IDPs + clinical covariates, in individuals with OCT images (N=55,469). (c) Bar plot of AUC values for single predictors and multimodal approaches, including PRS, IGSs, and clinical covariates, in individuals without imaging data (N=402,847). Error bars represent the standard errors (SE) of the AUC estimates.
