Machine learning models for drug–target interactions: current knowledge and future directions

Predicting the binding affinity between compounds and proteins with reasonable accuracy is crucial in drug discovery. Computational prediction of binding affinity between compounds and targets greatly enhances the probability of finding lead compounds by reducing the number of wet-lab experiments. Machine-learning and deep-learning techniques using ligand-based and target-based approaches have been used to predict binding affinities, thereby saving time and cost in drug discovery efforts. In this review, we discuss about machine-learning and deep-learning models used in virtual screening to improve drug–target interaction (DTI) prediction. We also highlight current knowledge and future directions to guide further development in this field.