Prediction of Drug-Disease Associations for Drug Repositioning Through Drug-miRNA-Disease Heterogeneous Network

• Main Authors: Hailin Chen, Zuping Zhang
• Format: Article
• Language: English
• Published: IEEE 2018-01-01
• Series: IEEE Access
• Subjects: Drug repositioning; target miRNAs; heterogeneous network; graph inference
• Online Access: https://ieeexplore.ieee.org/document/8421569/

Drug repositioning, which refers to the identification of new clinical indications for existing drugs, has become an important strategy for drug discovery. The most recent studies in pharmacogenomics have demonstrated that drugs can target microRNAs (miRNAs) and regulate their expression levels. Given the intriguing fact that the inappropriate expression of miRNAs is related to many kinds of human diseases, developing small-molecule drugs to target specific miRNAs and modulate their activities would be a promising approach to disease treatment, which offers an innovative insight for drug repositioning. In this paper, we proposed a miRNA-based computational method HNBI to infer novel drug-disease associations for drug repositioning. Similarity measurements and experimentally supported association information were first integrated to construct a three-layer drug-miRNA-disease heterogeneous network. Our method then updated the strength of weight between unlinked drug-miRNA, miRNA-disease, and drug-disease pairs iteratively till stabilized. Based on information flow on the heterogeneous network, the final weight of drug-disease associations was received by summarizing the values of paths connecting the two types of nodes. We prioritized the potential drug-disease associations according to the new weight. When applied to the collected data set for cross-validation experiments, our method showed superior performance in drug-disease association predictions compared with two state-of-the-art methods. Furthermore, our method HNBI incorporated information of target miRNAs to understand the mechanisms of action of drugs and the molecular mechanisms of diseases. A case study on the drug Terazosin indicated that some predicted indications with high ranks were supported by the recent literature, which further illustrated the practical usefulness of our method. Finally, comprehensive predictions of associations between drugs and diseases were released for future drug repositioning studies.