Application of similarity algorithms in drug-target prediction
Objective:It is a long and costly process from drug development to its clinical application.Using the similarity algorithm and machine learning algorithm to accurately predict drug-target interactions(DTI)can effectively assist drug research and development,and so forth improve efficiency and reduce costs.In order to provide a reference for the further study of drug-target prediction using the similarity algorithm,this paper makes a comprehensive analysis on the application of similarity algorithm in the study of drug-target prediction.Methods:The main literature retrieval platforms of China National Knowledge Infrastructure(CNKI),PubMed,Wanfang Data and VIP Network were used in this study.The similarity-based drug-target prediction research literature in the past ten years were collected and a database of literature on research subjects was established.We summarized the data sources,similarity construction algorithms of drugs and targets,and DTI prediction algorithm models in the drug-target prediction.Results:55 papers of high quality were collected,the results of which showed that the drug similarity is mainly constructed by the data of chemical structure,side effects,Anatomical Therapeutic Chemical(ATC)code and drug-target relationship.And target similarity is mainly constructed by the protein sequence and Gene Ontology(GO)annotation data.Different data can be measured by different similarity algorithms.Integrating multi-source similarity data can further improve the quality of the model.Drug-target interaction prediction algorithm models mainly include prediction models based on classification,network and matrix decomposition.Different prediction models have their respective advantages and disadvantages.Conclusion:It is concluded that drug-target prediction research has great developmental potential for by use of algorithm-assisted.It can help reduce the range of candidate drugs and target molecules,but some problems still remain unsolved.In the future,we need to further study this problem from two aspects:data optimization and algorithm model improvement.
NETL
NSTL
万方数据
