Similarity-Based Integrated Method for Predicting Drug-Disease Interactions

Abstract

The in silico prediction of potential interactions between drugs and disease is of core importance for effective drug development. Previous studies indicated that computational approaches for discovering novel indications of drugs by integrating information from multiple types have the potential to provide great insights to the complex relationships between drugs and diseases at a system level. However, each single data source is important in its own way and integrating data from different sources remains a challenging problem. In this article, we have proposed a new similarity combination method to integrate drug-disease association, drug chemical information, drug target domain information and target annotation information for drug repositioning. Specifically, we introduce interaction profiles of drugs (and of diseases) in a network, which are treated as label information and is used for model learning of new candidates. We compute multiple drugs and diseases similarity on these features, and use an integrated classifier for predicting drug-disease interactions. Comprehensive experimental results show that the proposed approach can serve as a useful tool in drug discovery to efficiently identify novel. Case studies show that our model has good performance.