HGIMDA: Heterogeneous graph inference for miRNA-disease association prediction

• Published in: Oncotarget Vol. 7; no. 40; pp. 65257 - 65269
• Main Authors: Chen, Xing, Yan, Chenggang Clarence, Zhang, Xu, You, Zhu-Hong, Huang, Yu-An, Yan, Gui-Ying
• Format: Journal Article
• Language: English
• Published: United States Impact Journals LLC 04.10.2016
• Subjects: Computational Biology - methods; Computer Simulation; Humans; MicroRNAs; Models, Theoretical; Neoplasms - genetics; Research Paper; microRNA; microRNA-disease association; disease; heterogeneous network; similarity
• ISSN: 1949-2553; 1949-2553
• DOI: 10.18632/oncotarget.11251

Recently, microRNAs (miRNAs) have drawn more and more attentions because accumulating experimental studies have indicated miRNA could play critical roles in multiple biological processes as well as the development and progression of human complex diseases. Using the huge number of known heterogeneous biological datasets to predict potential associations between miRNAs and diseases is an important topic in the field of biology, medicine, and bioinformatics. In this study, considering the limitations in the previous computational methods, we developed the computational model of Heterogeneous Graph Inference for MiRNA-Disease Association prediction (HGIMDA) to uncover potential miRNA-disease associations by integrating miRNA functional similarity, disease semantic similarity, Gaussian interaction profile kernel similarity, and experimentally verified miRNA-disease associations into a heterogeneous graph. HGIMDA obtained AUCs of 0.8781 and 0.8077 based on global and local leave-one-out cross validation, respectively. Furthermore, HGIMDA was applied to three important human cancers for performance evaluation. As a result, 90% (Colon Neoplasms), 88% (Esophageal Neoplasms) and 88% (Kidney Neoplasms) of top 50 predicted miRNAs are confirmed by recent experiment reports. Furthermore, HGIMDA could be effectively applied to new diseases and new miRNAs without any known associations, which overcome the important limitations of many previous computational models.