Chaotic Modified Binary Grey Wolf Optimizer Feature Selection method for Drug-Target interaction prediction Based on drug-protein Features

Drug-Target interaction prediction is a vital stage in drug development, in which there are lots of methods. Experimental methods that identify these relationship based on clinical remedies are timetaking, costly, laborions, complex, and with a lot of challenges. One of the new methods is computational methods. The development of new computational methods, which are more accurate, is better than the experimental methods in the aspect of cost and time. In this paper, to predict drug-target interaction, a new computational model is proposed that from the sequence of proteins information, different features such as Pse.AAC, PsePSSM, and drugs fingerprint information would be extracted, and then, these features are combined. In the next step, dine to a large number of features, the Chaotic MB-GWO (Chaotic Modified Binary Grey Wolf Optimizer) algorithm has been applied for feature selection. The final selected features are given to k-nearest neighbors (KNN) classification to predict the efficiency of this method. The accuracy of the (KNN) classifier on the golden standard datasets (Enzyme, ion channels (IC), G-protein-conpled receptors (GPCR), nuclear receptors (NR)) after feature selection is as follows: 97.18parcent, 97.71%,95.47%, and 93.05parcent. Referring to the results of experiments, this suggested method has an acceptable rate of drug-target interaction prediction and is compatible with the proposed methods in other papers.