Fig. 3
From: Inter-protein residue covariation information unravels physically interacting protein dimers
Prediction evaluation results for SVM models developed by multiple norm sets in series of ASA threshold groups in CCMpred and three distinct MI-based approaches. Multiple norm sets of ASA thresholds are employed to train SVM in 2 to 10 feature groups. Accuracy, sensitivity, specificity, precision, NPV, FPR, FNR and Informedness are, respectively, denoted by a, b, c, d, e, f, g, h. Note: 0.0–0.75 is a set of norm values for all ASA thresholds, ASA0.0 to ASA0.75. All other sets of ASA thresholds have similar symbols