System Prediction of Drug-Drug Interactions Through the Integration of Drug Phenotypic, Therapeutic, Structural, and Genomic Similarities

Abstract

Prediction of drug-drug interactions (DDIs) is an essential step in both drug development and clinical application. As the number of approved drugs increases, the number of potential DDIs rapidly rises. Several drugs have been withdrawn from the market due to DDI-related adverse drug reactions recently. Therefore, it is necessary to develop an accurate prediction tool that can identify potential DDIs during clinical trials. We propose a new methodology for DDIs prediction by integrating the drug-drug pair similarity, including drug phenotypic, therapeutic, structural, and genomic similarity. A large-scale study was conducted to predict 6946 known DDIs of 721 approved drugs. The area under the receiver operating characteristic curve of the integrated models is 0.953 as evaluated using five-fold cross-validation. Additionally, the integrated model is able to detect the biological effect produced by the DDI. Through the integration of drug phenotypic, therapeutic, structural, and genomic similarities, we demonstrated that the proposed method is simple, efficient, allows the uncovering DDIs in the drug development process and postmarketing surveillance.