The development of novel drugs is a time-consuming,labor-consuming,and costly process with the overall success rate no more than 10%.The prediction of drug-target interactions(DTIs)is fundamental for drug screening and drug repositioning.Accurate DTI prediction can significantly narrow down the screening of drug candidates and acceler-ate the drug discovery process.The traditional experimental method for identifying DTIs is tedious and expensive and accompanied by certain blindness,which restricts it from large-scale DTI identification.Recently,applying machine learning especially deep learning techniques to DTI prediction has become the mainstream.Although a series of meth-ods have been proposed in the last decade,DTI prediction is still a material-intensive and long-term work,and is chal-lenging to researchers.In this survey,we review literature related to DTI prediction,and summarize the methodologies,evaluation indicators,and data sources used in these works.We also analyze the shortcomings of existing works and propose future prospects.Our motivation is to help researches dedicated to drug discovery and development to have a comprehensive understanding on the latest progress of DTI prediction so as to improve their research efficiency and re-search quality.
万方数据