Using Supervised Machine Learning Algorithms For Drug Target Interaction Prediction

Abstract

In conjunction with the advent of bioinformatics, the evolution occurred in the field of drug discovery worldwide. In the literature numerous machine learning (ML), deep learning, and graph theory approaches have been implemented for drug discovery tasks. However, the drug-target interaction (DTI) problem for the existing and new drugs has been anticipated for a long time in drug repurposing and drug discovery areas. The proposed study predicts the drug-target interactions and to achieve this, a computational pipeline has been developed for the heterogeneous network. Firstly, this study obtains the low dimensional vector for the graphical nodes using the node2vec method. Secondly, multiple machine learning methods have been applied to predict drug-target interactions. Lastly, for the evaluation of the proposed models, we calculated the AUROC and AUPRC values and the obtained results indicate that non-linear SVM and Logistic regression performed better than other models with the AUROC, and AUPRC values of 0.8317 and 0.8260 respectively.