“…With the fast expansion of resolved sequences and structural data of proteins, several kinds of computational methods such as molecular dynamics methods [17,18] and machine learning methods have been proposed to predict PPI sites. Among these methods, machine-learning methods are most successful, using the following models [14,19]: support vector machine (SVM) [20][21][22][23][24][25] and fuzzy SVM [26], neural networks (NN) [27][28][29][30][31][32], Bayesian networks (BN) [33,34], naive Bayes classifier (NBC) [35,36], random forests (RF) [12,[37][38][39], cascade random forests (CRF) [40], conditional random fields (CRF) [41], extreme learning machine (ELM) [42], L1-logreg classifier [43], and the ensemble method [14,15,44,45]. As a recent development of neural networks, deep-learning is a rapidly growing branch of machine learning and has also been used to predict PPI sites [19].…”