Predicting drug-target interactions via sparse learning.

Drug-target interaction (DTI) prediction plays a very important role in drug development. Biochemical experiments or in vitro methods to identify such interactions are very expensive, laborious and time-consuming. Therefore, in silico approaches including docking simulation and machine learning have been proposed to solve this problem. In particular, machine learning approaches have attracted increasing attentions recently. However, in addition to the known drug-target interactions, most of the machine learning methods require extra information such as chemical structures, genome sequences, binding types and so on. Whenever such information is not available, they may perform poor. Very recently, the similarity-based link prediction methods were extended to bipartite networks, which can be applied to solve the DTI prediction problem by using topological information only. In this work, we propose a sparse learning method to solve the DTI prediction problem, which does not require extra information and performs much better than similarity-based methods. We compare the proposed method with similarity-based methods including common neighbor index, Katz index and Jaccard index on the DTI prediction problem over the four renowned and benchmark datasets. The proposed method performs remarkably better. The results suggest that although the proposed method utilizes only the known drug-target interactions, it performs very satisfactorily. The method is very suitable to predict the potential uses of the existing drugs, especially, when extra information about the drugs and targets is not available.