Identification of Small Molecule–miRNA Associations with Graph Regularization Techniques in Heterogeneous Networks
Cong Shen,Jiawei Luo,Wenjue Ouyang,Pingjian Ding,Hao Wu
DOI: https://doi.org/10.1021/acs.jcim.0c00975
IF: 6.162
2020-11-09
Journal of Chemical Information and Modeling
Abstract:MicroRNAs (miRNAs) are significant regulators of post-transcriptional levels and have been confirmed to be targeted by small molecule (SM) drugs. It is a novel insight to treat human diseases and accelerate drug discovery by targeting miRNA with small molecules. Computational approaches for discovering novel small molecule–miRNA associations by integrating more heterogeneous network information provide a new idea for the multiple node association prediction between small molecule–miRNA and small molecule–disease associations at a system level. In this study, we proposed a new computational model based on graph regularization techniques in heterogeneous networks, called identification of small molecule–miRNA associations with graph regularization techniques (SMMARTs), to discover potential small molecule–miRNA associations. The novelty of the model lies in the fact that the association score of a small molecule–miRNA pair is calculated by an iterative method in heterogeneous networks that incorporates small molecule–disease associations and miRNA–disease associations. The experimental results indicate that SMMART has better performance than several state-of-the-art methods in inferring small molecule–miRNA associations. Case studies further illustrate the effectiveness of SMMART for small molecule–miRNA association prediction.The Supporting Information is available free of charge at <a class="ext-link" href="/doi/10.1021/acs.jcim.0c00975?goto=supporting-info">https://pubs.acs.org/doi/10.1021/acs.jcim.0c00975</a>.Baseline method; details of association data and similarity data (Table S1); average recall, specificity, and <i>G</i>_mean of four methods within top <i>k</i> cutoffs (Table S2); parameter sensitivity analysis (Figure S1); comparison of SMMART with the other methods based on 80%, 70%, 60%, and 50% known associations (Figures S2–S5); performance comparison between SMMART and SMMART*: (a) recall of SMMART and SMMART* within top 5–30% and (b) <i>G</i>_mean of SMMART and SMMART* within top 5–30% (Figure S6); average recall, specificity, <i>G</i>_mean, and precision of the six methods based on 90%, 80%, 70%, 60%, and 50% known associations within top <i>k</i> cutoffs (Table S3–S7); average recall, specificity, <i>G</i>_mean, and precision based on association data at different top <i>k</i> cutoffs (Table S8); top 10 potential disease candidates discovered by SMMART for the two selected diseases (Table S9) (<a class="ext-link" href="/doi/suppl/10.1021/acs.jcim.0c00975/suppl_file/ci0c00975_si_001.pdf">PDF</a>)This article has not yet been cited by other publications.
chemistry, multidisciplinary, medicinal,computer science, interdisciplinary applications, information systems