Prediction Of The Ncrnas-Associated With Cancer And Biomarkers Mediated By Ncrnas

In organisms,the transcriptional RNA molecules can broadly be divided into two categories:coding RNAs(mRNA)and non-coding RNA(ncRNA)based on whether they directly guide and participate in the synthetics of proteins(or peptides)or not.At first,it was generally believed that most of ncRNAs were "junk RNAs".With the study of the mechanism and function of ncRNA,it has been found that ncRNA is involved in many important life activities,such as DNA replication,RNA splicing,translation,genome defense and so on.Through the study of the human genome and the mechanism of diseases,it is found that not only genes are associated with the occurrence and development of human complex diseases,but also dysregulated or aberrant ncRNA can result in various complex diseases.Therefore,the prediction of associations between ncRNA and disease and the identification of disease-associated biomarkers are important for understanding the functions of ncRNA as well as the prevention and treatment of disease.miRNA is an important endogenous regulator of ncRNA in eukaryotes.The mature miRNA combines with the untranslated region of 3'end in a complementary(complete or incomplete)paring method,which inhibits the translation of the target mRNA and silences the expression of the genes.It has been shown that miRNA was involved in a series of important life processes,such as early development,virus defense,cell proliferation,hematopoiesis processes,and cell apoptosis.The aberrant regulation of miRNA is an important cause of cancer and other complex diseases,so the prediction of disease-related miRNA is an important direction for understanding of the causes of disease,and it contributes to the prevention,diagnosis and treatment of complex disease.However,it is difficult to identify and verify the disease related miRNA by biological experiments,because the different dissolution temperatures of miRNA short sequences and the similarity of the sequence in the miRNA family,which makes the experimental results more deviant and even wrong.Moreover,in the face of massive miRNA data,identifying the relationship between miRNA and disease by biological experiment will cost a lot of manpower and material resources.Therefore,based on the above problems,we utilize the known associations between miRNA and cancer to predict the associations which have not yet been confirmed by experiments between miRNA and cancer by the similarity of miRNAs and cancers.Considering miRNA mainly play roles by combining with 3'UTR region of target genes,so we predict the potential target genes of miRNA by reliable prediction tools-TargetScan and miRanda.Next,we calculate the similarity of miRNA by utilizing the overlapping of targets.For cancer,we obtain the semantic similarity matrix of cancer through the Medical Subject Headings(MeSH).Finally,we construct a predictive regression model based on adjacent matrix from the known relationship among miRNA and cancer,cancer-cancer network and miRNA-miRNA functional similarity network and predict the miRNA associated with 134 types of cancer.The main results are as follows:1.Through the case study of colon cancer,stomach cancer,esophageal cancer and liver cancer,it is found that 45,41,39 and 41 miRNAs in the top 50 of concordance scores have been confirmed to be associated with the corresponding cancer,the unconfirmed 5,9,11 and 9 miRNAs of 4 cancers are thought to be the potential miRNAs associated with cancer.These results show that the proposed prediction model not only can effectively identify the cancer-related miRNA,but also can predict the potential cancer-related miRNA.2.Through functional analysis of miRNA with the larger degree(degree:the number of cancers related to these miRNAs in the predicted relationship),it is found that some miRNAs have large degree,and they are cancer-related important miRNAs,such as mir-664a and mir-4789.Studies have shown that mir-664a is known to be related to 34 cancers and is an important regulator of cancer.Moreover,through analyzing the KEGG pathway of the target genes of mir-4789,it is found that the target genes are significantly enriched in the cancer pathways.The above results indicate that the larger the degree of predicted miRNA is,the more likely it is to be associated with multiple cancers.Studies have shown that the dysregulation or aberrant expression of miRNA and lncRNA can result in complex diseases,and lncRNA is important participant of ceRNA mechanism,which can mislead miRNAs away from their natural targets as a competitor and reduce the regulatory effect on its targets.The current research is mainly concerned with the recognition of biomarkers of single molecules,but the occurrence of complex diseases is often due to the dysfunction of relevant regulatory network,rather than the malfunction of individual molecules.Hence,based on the current research on non-coding RNA and mechanism of disease,this paper aims to predict the biomarkers of miRNA-lncRNA interactions related to cancer.In this paper,the 155653 experimental confirmed miRNA-lncRNA interactions are selected as the basic interactions.We construct the individual-special miRNA-lncRNA regulatory network(ISMLN)by utilizing the expression profiles of 1046 miRNAs and 12727 lncRNAs and obtain a basic miRNA-lncRNA network(BMLN)by the ISMLNs of all patients.Next,we obtain the candidate miRNA-lncRNA pairs based on the Significance Score by analyzing whether the 155653 miRNA-lncRNA pairs significantly change in the basic miRNA-lncRNA network.In order to further screen potential miRNA-lncRNA biomarkers,we select the candidate miRNA-lncRNA pairs as features,the difference of Pearson correlation coefficients between non-tumor samples and tumor samples as the values of features.We apply the Random Forest(RF)to distinguish tumor samples from non-tumor samples and filter the potential miRNA-lncRNA biomarkers based on the Features Contribution of candidates.The miRNA-lncRNA biomarkers of breast cancer are identified in this paper,the main results are as follows:1.Comparing the classification results of node biomarkers(miRNA and lncRNA)with edge biomarkers(miRNA-lncRNA).It is found that when we take the top 5 of this two kinds of biomarkers,the average accuracy of miRNA biomarkers is 96.26%,the maximum accuracy is 98.32%.The average accuracy of lncRNA biomarkers is 98.09%,and the maximum accuracy is 99.42%.For miRNA-lncRNA edge biomarkers,the average accuracy is 99.23%,and the maximum accuracy is 99.96%.It is obvious that the accuracy of these two kinds of biomarkers has little difference when we take the front biomarkers,and the edge biomarkers have a little advantage.2.We further select the last 5 biomarkers as the features for classification.The accuracy of miRNA biomarkers is 40.16%,and the accuracy of lncRNA biomarkers is 54.05%,but the accuracy of miRNA-lncRNA edge biomarkers still can achieve 91.26%.So the selected miRNA-lncRNA edge biomarkers contain more bioinformatics information,and have more stability and accuracy than node biomarkers.3.By analyzing the changes in the expression level of miRNA and lncRNA in the potential biomarkers.It is found that proposed method can identify non-differentially expressed biomarkers.In addition,the specific biomarkers among different cancers can be identified by this method.