WGMFDDA: A Novel Weighted-Based Graph Regularized Matrix Factorization for Predicting Drug-Disease Associations

• Published in: Intelligent Computing Methodologies pp. 542 - 551
• Main Authors: Wang, Mei-Neng, You, Zhu-Hong, Li, Li-Ping, Chen, Zhan-Heng, Xie, Xue-Jun
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Drug-disease association; Graph regularization; Matrix factorization; nearest neighbor
• ISBN: 303060795X; 9783030607951
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-60796-8_47

Identification of drug-disease associations play an important role for expediting drug development. In comparison with biological experiments for drug repositioning, computational methods may reduce costs and shorten the development cycle. Thus, a number of computational approaches have been proposed for drug repositioning recently. In this study, we develop a novel computational model WGMFDDA to infer potential drug-disease association using weighted graph regularized matrix factorization (WGMF). Firstly, the disease similarity and drug similarity are calculated on the basis of the medical description information of diseases and chemical structures of drugs, respectively. Then, weighted K\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ K $$\end{document}-nearest neighbor is implemented to reformulate the drug-disease association adjacency matrix. Finally, the framework of graph regularized matrix factorization is utilized to reveal unknown associations of drug with disease. To evaluate prediction performance of the proposed WGMFDDA method, ten-fold cross-validation is performed on Fdataset. WGMFDDA achieves a high AUC value of 0.939. Experiment results show that the proposed method can be used as an efficient tool in the field of drug-disease association prediction, and can provide valuable information for relevant biomedical research.