Identification Of Cancer-related MiRNAs Based On MiRNA-gene Association

With the higher prevalence and the increasing death toll of cancer in China,the prevention and treatment of cancer has become an urgent problem to be solved in our country.Related studies have shown that a class of single-stranded non-coding RNA(MicroRNAs,miRNA)plays an important role in the occurrence and development of cancer.Therefore,the identification of cancer-related miRNA is the key step for studing the specific mechanism of miRNA in cancer and finding the diagnostic markers and treatment targets of cancer.The methods that are based on the analysis and computation of bioinformatics have the characteristics of low cost and high efficiency,which is always superior to the traditional biological experiments.Currently,some researchers have proposed many computational methods for identifying cancer-related microRNAs,which have many problems.For example,(1)The differential expression analysis leads to a very high false positive rate.(2)The methods are based on the similarity of miRNAs and diseases,which needs to be further studied.(3)Some methods depend on the trained classification model.However,these methods were not considered the lack of negative samples.(4)It did not consider a variety of influencing factors and the huge scale of calculation that some methods analysed the association of miRNAs and genes.Based on the analysis of existing methods,this paper focuses on the identification methods of cancerrelated miRNAs and proposes three methods to identify cancer-related miRNAs.The main work is as follows:(1)An identification method integrating four data sources(genes,proteins,miRNAs and driving genes)is proposed.This method effectively combines the biological characteristics corresponding to these four data sets.And it considers non-coding genes in the consolidation process of genes and proteins datas.For the first time,driving genes are introduced into the identification of cancer-related miRNAs.The experimental results show that this method has better recognition effect and higher efficiency than the old method.(2)The method based on miRNA-gene regulatory subnetwork is proposed.By integrating gene and miRNA expression data and miRNA-target gene data,the problems caused by the tissue specificity of miRNA and the low applicability caused by a single data were effectively solved.The fuzzy clustering is used to divide the whole miRNA-gene regulatory network,which solves the problems of excessive computing scale and complex redundant data.The experimental results,on four actual cancer data sets,show this method has significant advantages compared with existing methods.(3)A novel identification method considered the abnormal pathways is proposed.On the one hand,the advantages of the two data are combined by integrating genes,miRNAs expression data and miRNA-target genes data.In the other hand,the differential expression analysis of genes is used to identify the abnormal pathways.The association between miRNAs and abnormal pathways is obtained from the correlation analysis of miRNAs target genes and genes in abnormal pathway.Then,the cancer-related miRNAs are further identified.The experimental results show that the proposed method has better identification effect on a variety of evaluating indicator than the old method based on the enrichment of target genes in abnormal pathways and other methods.