Analysis and prediction of ion channel inhibitors by using feature selection and Chou's general pseudo amino acid composition

Venomous animals produce toxins that inhibit ion channels with high affinity. These small peptide inhibitors are used in the characterization of ion channels structurally as well as pharmacologically. So, identification of these toxins is an important task. In this study, based on the pseudo amino acid (PseAA) composition and feature selection method, the random forest algorithm was used for predicting three different groups of ion channel inhibitors. The prediction results indicated that our algorithm achieved the sensitivity of 60.00% for calcium channel inhibitor, 71.90% for potassium channel inhibitor and 86.80% for sodium channel inhibitor when evaluated by the jackknife test. In addition, for comparing with other algorithms, this algorithm was used to predict the dataset with 343 ion channel inhibitors, and the higher predictive success rates than the previous algorithms were obtained by our algorithm.