Predicting Comprehensive Drug-Drug Interactions for New Drugs via Triple Matrix Factorization

There is an urgent need to discover or deduce drug-drug interactions (DDIs), which would cause serious adverse drug reactions. However, preclinical detection of DDIs bears a high cost. Machine learning-based computational approaches can be the assistance of experimental approaches. Utilizing pre-market drug properties (e.g. side effects), they are able to predict DDIs on a large scale before drugs enter the market. However, none of them can predict comprehensive DDIs, including enhancive and degressive DDIs, though it is important to know whether the interaction increases or decreases the behavior of the interacting drugs before making a co-prescription. Furthermore, existing computational approaches focus on predicting DDIs for new drugs that have none of existing interactions. However, none of them can predict DDIs among those new drugs. To address these issues, we first build a comprehensive dataset of DDIs, which contains both enhancive and degressive DDIs, and the side effects of the involving drugs in DDIs. Then we propose an algorithm of Triple Matrix Factorization and design a Unified Framework of DDI prediction based on it (TMFUF). The proposed approach is able to predict not only conventional binary DDIs but also comprehensive DDIs. Moreover, it provides a unified solution for the scenario that predicting potential DDIs for newly given drugs (having no known interaction at all), as well as the scenario that predicting potential DDIs among these new drugs. Finally, the experiments demonstrate that TMFUF is significantly superior to three state-of-the-art approaches in the conventional binary DDI prediction and also shows an acceptable performance in the comprehensive DDI prediction.