Fig. 4
From: Decrypting orphan GPCR drug discovery via multitask learning
The mRMR-based feature selection. A The validation performance for top-ranked feature selection model sets. B Distribution analysis of 162 aligned protein features of top-ranked 200 features from the feature-selected model (MTL-AG-ATG-FS). C The scatter plot of pairwise Tanimoto similarity (Ts) based on selected protein features calculated as individual parts or full-length sequence plotting against validation performance on the MTL-AG-ATG-FS model