Modeling relational drug-target-disease interactions via tensor factorization with multiple web sources

Abstract

Modeling the behaviors of drug-target-disease interactions is crucial in the early stage of drug discovery and holds great promise for precision medicine and personalized treatments. The growing availability of new types of data on the internet brings great opportunity of learning a more comprehensive relationship among drugs, targets, and diseases. However, existing methods often consider drug-target interactions or drug-disease interactions separately, which ignores the dependencies among these three entities. Also, many of them cannot directly incorporate rich heterogeneous information from diverse sources. In this work, we investigate the utility of tensor factorization to model the relationships of drug-target-disease, specifically leveraging different types of online data. Our motivation is two-fold. First, in human metabolic systems, many drugs interact with protein targets in cells to modulate target activities, which in turn alter biological pathways to promote healthy functions and to treat diseases. Instead of binary relationships of or, a tighter triple relationships should be exploited to better understand drug mechanism of actions (MoAs). Second, medical data could be collected from different sources (i.e., drug's chemical structure, target's sequence, or expression measurements). Therefore, effectively exploiting the complementarity among multiple sources is of great importance. Our method elegantly explores a tensor together with complementarity among different data sources, thus improves prediction accuracy. We achieve this goal by formulating the problem into a coupled tensor-matrix factorization problem and directly optimize it on the nonlinear manifold. Experimental results on real-world datasets show that the proposed model outperforms several competitive methods. Our model opens up opportunities to use large Web data to predict drugs' MoAs in pharmacological studies.