A Review on Predicting Drug Target Interactions Based on Machine Learning.

The prediction of drug-target interactions (DTIs) is a key preliminary step for drug discovery and development due to the high risk of failure as well as the long validation period of in vitro and in vivo experiments. Nowadays, with the swiftly growing power in solving scientific problems, machine learning has become an important tool in DTI prediction. By simply categorizing them into traditional machine learning-based approaches and deep learning-based ones, this review discusses some representative approaches in each branch. After a brief introduction on traditional methods, we firstly pay large attention to the data representation of deep learning-based methods, which can be summarized with 5 different representations for drugs and 4 for proteins. Then we introduce a new taxonomy of deep neural network models for DTI prediction. Furthermore, the commonly used datasets and evaluation metrics were also summarized for an easier hands-on practice.