The Research On Disease-related MiRNAs Prediction Methods And Its Applications Based On Similarity Network

Ribonucleic acid(RNA)is a genetic information carrier existing in biological cells,some viruses and viroids.There are coding RNAs that encode proteins in RNA,and non-coding RNAs(ncRNA)that do not encode proteins.Non-coding RNAs account for approximately 98% of the entire genome according to the genome-wide transcriptional studies.These ever-growing list of ncRNAs has been found to play important roles in biological and pathological processes and their aberrant expression or mutation always contributes to a variety of diseases.microRNA(miRNA)is small endogenous ncRNA of about 22 nucleotides long,that functions in RNA silencing,post-transcriptional regulation of gene expression and exerting inhibitory effects at the transcriptional level by binding to the 3'UTR or 5'-end of a specific target gene by means of complete or incomplete based-pairing principles.miRNA has a wide range of gene expression and regulation,and is involved in biological processes such as individual growth and development,cell proliferation and differentiation,and apoptosis.miRNA play an important role in the process of the generation and development of human diseases such as cardiovascular and cerebrovascular diseases,infectious diseases,malignant tumors and metabolic disorders.Accumulating studies have indicated that some miRNA expressions differ in the expression profiles of cancerous and paracancerous tissues,and that miRNA expression profiles are distinct for each stage of the tumor or even cancers.Some miRNAs can be markers of cancer detection,or may be targets for disease treatment or as predictors of disease efficacy.The close relationship between miRNAs and diseases is closely related to the identification of miRNAs and their association with diseases,which is of great significance for early detection,diagnosis,treatment and prognosis of diseases.However,the study of various disease-related miRNAs is only an initial phase,although the continual disease-related miRNAs are time-consuming and costly found by biological experiments.This paper explores the relationship between miRNA and disease from the perspective of bioinformatics for finding the miRNAs with the most relevant potential of disease,hoping to provide the best experimental targets for biological experiments.The main contents are as follows:(1)First,the characteristics,functions,and related biomolecular networks of miRNAs are studied.Then,this paper studies the existing database resources related to miRNAs,the existing methods for predicting disease miRNAs,and analyzes the prediction basis,advantages and disadvantages of these methods.(2)A network similarity integeration method for predicting miRNA-disease associations.There are some limitations of existing computational methods for predicting miRNA-disease associations,suc as low prediction accuracy,negative samples,insufficient performance of leave-one-out cross validation(LOOCV),long validation time,no prediction of isolated disease without miRNAs.In this study,we develop a network similarity intergration method(NSIM)for discovery potential disease-related miRNAs.NSIM calculates the potential miRNA-disease association scores by integrating the miRNA and disease vector space score,which are calculate using cosine similarity.We evaluate the NSIM using leave-one-out cross validation.The area under the curve of the method is 0.9475,indicating its outstanding performance.Case studies on prostate,breast,and colon neoplasms further proved the outstanding performance of the NSIM to predict not only disease–related miRNAs but also isolated diseases(diseases without any related miRNAs).(3)SRMDAP: SimRank and density-based clustering recommender model for miRNA-disease association prediction.This paper propose a novel computational method based on SimRank and density-based clustering recommender model for miRNA-disease association prediction(SRMDAP)to solve these limitations.The SRMDAP constructs miRNA similarity subnetwork by using SimRank to calculate network topological similarity between miRNAs based on miRNA-mRNA(message RNA)interaction network.The disease similarity subnetwork is similar to miRNA similarity subnetwork and based on the disease-gene network.Then,using densitybased clustering recommender model,the SRMDAP integrates miRNA similarity subnetwork,disease similarity subnetwork,and experimentally verified miRNAdisease associations to predict potential associations between miRNAs and diseases.In this work,LOOCV experiment and case studies about two important cancers(kidney neoplasms and colorectal neoplasms)have indicated the excellent predictive performance of SRMDAP.The SRMDAP can also predict isolated diseases and isolated miRNAs.(4)Using miRNA family information and cluster information to improve predict accuracy of disease related miRNAs.The family information and cluster information of miRNAs are not considered in many existing miRNA-disease association prediction methods.Sequences(especially seed sequences)of highly homologous miRNAs are classified as a miRNA family.The miRNAs in the same miRNA family are not always adjacent in location.The same miRNA family members have similar functions.miRNA clusters are miRNAs that are very close together on the genome and are often coexpressed.To further improve the accuracy of miRNA-disease association prediction,this paper propose using miRNA family and cluster information to improve predict accuracy of disease related miRNAs(FCMDAP)method.In the FCMDAP method,the similarity between miRNAs is calculated based on miRNA-mRNA interactions using mutual information in miRNA similarity calculation,and miRNA family information are added to reconstruct the miRNA similarity network.Then FCMDAP use the neighbors information of the miRNA and of the same miRNA cluster to predict potentioal scores are calculated,which is based on the miRNA similairy network and known miRNA-disease associations network.In disease similarity calculation,the similarity between diseases is also calculated based on the known disease-gene interaction information using mutual information,and the disease similarity network is constructed by adding the semantic similarity indicated by the disease DAG.The miRNAs associated with the disease are predicted separately in the miRNA similarity cyberspace and the disease similarity cyberspace based on the k nearest neighbor recommendation algorithm,then they are integrated together.In this work,LOOCV experiment and case studies about colorectal,prostate,and pancreative neoplasms have indicated the excellent predictive performance of FCMDAP.The FCMDAP can also predict isolated diseases and isolated miRNAs.