Prediction Of MiRNA-disease Association Based On Multi-source Data Fusio

Micro RNA(miRNA)is a kind of non-coding small single-stranded RNA molecule in animals and plants,which regulates the expression of target gene m RNA.Mi RNA is involved in many life processes of human beings and the regulation of miRNA on gene expression may lead to some diseases.In traditional validation experiments,it usually studies the correlation between miRNA and disease by biological methods,but this method has the disadvantages of long experimental period and high cost.In the heterogeneous networks of the current bioinformatics prediction methods,the prediction results are not good enough because of the sparse matrix and insufficient data source information.To solve these problems,this paper adopts multi-source data to construct high-quality complex network,and proposes a prediction model based on matrix completion and a prediction model based on gene ontology data.The specific work includes the following three aspects:(1)A matrix completion based miRNA-disease association prediction model MDFMC was proposed.In this model,the constructed miRNA and disease similarity matrix is processed by matrix completion,and the completed similarity matrix is integrated into the miRNA network and disease network construction.Label propagation algorithm was used to predict in complex networks with miRNA subnets and disease subnets respectively.To test the predictive performance of the MDFMC model,three cross-validation methods were used to compare the model with related models,and specific disease cases were used to evaluate the predictive results of the model.Experimental results show that MDFMC model has good predictive performance.(2)A miRNA targeted similarity measuring method using Gene Ontology data was designed.Based on this method,a miRNA-disease association prediction model GOBMDA was proposed.In this model,the target genes of miRNA are mapped to Gene Ontology data to construct the miRNA targeted similarity matrix.A similarity network between miRNA and disease was constructed using multiple similarity matrices from different data sources,and a random walk algorithm with restart was used for prediction in the heterogeneous network.Finally,We used cross validation methods and case analysis methods to illustrate the prediction effectiveness of GOBMDA model.(3)An online prediction platform based on Springboot framework was designed.It provide relevant researchers with prediction services for miRNA-human disease association based on MDFMC model and GOBMDA model,and it is convenient to evaluate of the validity of the model.The test results show that the prediction platform has good human-machine interaction and quick processing response.