The Research On Small Molecule Drug-miRNA Associations And Virus-drug Associations Prediction

The mechanism of human complex diseases has always been a challenge in medicine and biology.At present,with the development of medical research,researchers have found that overexpression or underexpression of some miRNAs can lead to pathological changes in normal cells.These miRNAs have been reported to be drug targets of complex diseases.In December 2019,the outbreak of COVID-2019 produced a severely burden to the medical and health system.Researchers have tried to use drugs,treat other viral pneumonia,to treat COVID-2019.Traditional methods,requires a lot of time and resources,are used to predict a new miRNA-targeted drug.Drug repositioning is aimed at discovering new drug properties of drugs already on the market and provides a realistic and feasible way for the development of targeted miRNA drugs and anti-coronavirus drugs.At present,researchers have developed computational models for mining potential miRNA-targeted drugs.It is a research hotspot that how to integrate a variety of biological data to obtain better prediction performance.The main research work of this paper is as follows:1.The computational model(RWNS)was proposed to predict small molecule drug-miRNA associations(SMiR).The prediction process can be divided into the following three steps.First,RWNS constructs three association networks(small molecule drug-miRNA association,diseasemiRNA association,small molecule drug-disease association)and three similarity matrices(small molecule drug similarity matrix,miRNA similarity matrix and disease similarity matrix).Second,RWNS bults a credible negative sample screening framework.Finally,an unbalanced random walk-based model,similarity matrices,association networks and negative sample screening framework are integrated into a unified framework to discover potential SMiR.Experimental results and case study show that RWNS can effectively mine the potential small moleculemiRNA drug associations.2.The unbalanced bi-random walk model(VDA-UBiRW)was designed to screen anti-coronavirus drugs.VDA-UBiRW computed virus genetic sequence similarity,medicinal chemical structure similarity and virus-small molecule associations matrix.Second,unbalanced random walks are used to explore potential anti-coronavirus drugs.Finally,the performance of VDA-UBiRW was verified through Experimental results and molecular docking experiments.3.The SMiR-GCNN,a computational model based on deep learning,was developed to predict the associations of small molecule dug and miRNA.SMiR-GCNN uses graph neural networks to extract sub-graph vectors of the spatial structure data of small molecule drugs.Second,convolutional neural networks to extract sub-sequence features of miRNA sequence data.Finally,the "neural attention mechanism" is used to speculate on potential SMiR.Experimental results and case study show that SMiR-GCNN can obtain better prediction performance.