A Network Embedding Based Approach to Drug-Target Interaction Prediction Using Additional Implicit Networks

Identifying novel drug-target interactions (DTIs) is a crucial step in drug discovery. Since experimentally determining DTIs is expensive and time-consuming, it becomes popular to employ computational methods for providing promising candidate DTIs. However, in the existing computational methods, the drug implicit network and target implicit network constructed from a DTI network (a bipartite network) have been ignored in the DTI prediction problem, while such implicit networks constructed from a bipartite network have been proven useful in other problems, e.g., the link prediction task in a bipartite network. Motivated by that, we propose a novel DTI prediction method which considers the implicit networks in addition to drug structure similarity network and target sequence similarity network. The experiments over five real-world DTI datasets demonstrate the competitive performance of the proposed method compared to the state-of-the-art methods. The code is available at https://github.com/BrisksHan/NE-DTIP.