In Silico Prediction of Small Molecule-miRNA Associations Based on the HeteSim Algorithm

• Published in: Molecular therapy. Nucleic acids Vol. 14; pp. 274 - 286
• Main Authors: Qu, Jia, Chen, Xing, Sun, Ya-Zhou, Zhao, Yan, Cai, Shu-Bin, Ming, Zhong, You, Zhu-Hong, Li, Jian-Qiang
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01-03-2019; Elsevier Limited; American Society of Gene & Cell Therapy; Elsevier
• Subjects: association prediction; Cancer; Computer applications; Datasets; Genes; heterogeneous network; HeteSim algorithm; Identification; Ligands; Medical research; microRNA; MicroRNAs; miRNA; Neuromuscular diseases; Semantics; small molecule; Statistical analysis; microRNA; association prediction; small molecule; heterogeneous network; HeteSim algorithm
• ISSN: 2162-2531; 2162-2531
• DOI: 10.1016/j.omtn.2018.12.002

Targeting microRNAs (miRNAs) with drug small molecules (SMs) is a new treatment method for many human complex diseases. Unsurprisingly, identification of potential miRNA-SM associations is helpful for pharmaceutical engineering and disease therapy in the field of medical research. In this paper, we developed a novel computational model of HeteSim-based inference for SM-miRNA Association prediction (HSSMMA) by implementing a path-based measurement method of HeteSim on a heterogeneous network combined with known miRNA-SM associations, integrated miRNA similarity, and integrated SM similarity. Through considering paths from an SM to a miRNA in the heterogeneous network, the model can capture the semantics information under each path and predict potential miRNA-SM associations based on all the considered paths. We performed global, miRNA-fixed local and SM-fixed local leave one out cross validation (LOOCV) as well as 5-fold cross validation based on the dataset of known miRNA-SM associations to evaluate the prediction performance of our approach. The results showed that HSSMMA gained the corresponding areas under the receiver operating characteristic (ROC) curve (AUCs) of 0.9913, 0.9902, 0.7989, and 0.9910 ± 0.0004 based on dataset 1 and AUCs of 0.7401, 0.8466, 0.6149, and 0.7451 ± 0.0054 based on dataset 2, respectively. In case studies, 2 of the top 10 and 13 of the top 50 predicted potential miRNA-SM associations were confirmed by published literature. We further implemented case studies to test whether HSSMMA was effective for new SMs without any known related miRNAs. The results from cross validation and case studies showed that HSSMMA could be a useful prediction tool for the identification of potential miRNA-SM associations.