Research on Drug-Target Interactions Prediction: Network similarity-based approaches

Abstract

Accurate identification of drug-target interactions can help researchers shorten the time of new drug development and reduce the blindness and cost of new drug research. For this purpose, the link prediction technology is used in this paper to predict the accuracy of 24 kinds of similarity indexes in the Matador database. The accuracy of link prediction based on structural similarity depends on whether the definition of structural similarity can grasp the structural characteristics of the target network well. And it does not need to know the information of nodes and edges of the network in advance. The results show that compared with other algorithms, the accuracy of Local Random Walk (LRW) model is the highest when the step size is 5, and the algorithm with the best area under the ROC curve ( AUC ) stability is Hub Promoted Index (HPI).