Fig. 1From: Inter-protein residue covariation information unravels physically interacting protein dimersContribution of features to SVM classifications. a For each single ASA threshold, SVM is trained by norm values of residue contact matrices as a single input feature for further prediction. b SVM models are developed for multiple groups norm values in different sets of ASA thresholds to predict the possible interaction in each heterodimeric pairBack to article page