The Research On Methods Of Predicting MiRNA-target Associations By Integrating Multiple Types Of Genomic Data

MicroRNA(miRNA)are short noncoding RNAs that play important roles in regulating gene expressing.The perturbed miRNAs often cause the unnormal of the biological process and regulation pathways as they have effects on their target mRNA.In addition,miRNA and its target are involved in different diseases.Therefore,deciphering miRNA targets is crucial for discovering miRNA function and regulatory mechanisms,and also important for finding potential therapeutics for complex diseases.With the rapid growth and wide use of high throughout techniques,a large volume of omics data has been generated,which provides new opportunities for studying the complex relationships among many kinds of biological molecules based on network.Therefore,how to use the multiple types of genomic data to identify miRNA-target association is the research focus in bioinformatics.Here,we propose two heterogeneous network-based approaches to identify miRNA-target interactions.To use the validated miRNA regulatory relationship effectively,we proposed network-based approaches,RMLM and RMLMSe,to predict miRNA-target associations based on meth-path.RMLM and RMLMSe can reconstruct the missing associations for all the miRNA-target in one or more diseases network simultaneously and RMLMse demonstrates that the integration of sequence information can improve the performance of RMLM.In RMLM,we use Relatedness Measure(RM)to evaluate different relatedness between miRNA and its target based on different meta-paths;Logistic regression and maximum likelihood estimation(MLE)are employed to estimate the weight of different meta-paths.In RMLMSe,sequence information is utilized to improve the performance of RMLM.Here,we carry on fivefold cross validation and pathway enrichment analysis on four networks to prove the performance of our methods.The fivefold experiments show that our methods have higher AUC scores compared with other methods and the integration of sequence information can improve the performance of the miRNA-target association prediction.The results of pathway enrichment analysis indicate that RMLM and RMLMSe are reasonable and credible.Although the better performance is obtained by RMLM on the whole,the predictive results should be further improved for the small output.A method called GLRWR is proposed to predict miRNA target.Firstly,GLRWR constructe gaussian interaction profile kernel similarity for miRNA-miRNA,gene-gene from known miRNA-gene,gene-miRNA network respectively and then construct integrated miRNA similarity matrix,gene similarity matrix based on miRNA functional similarity,gene similarity and gaussian interaction profile kernel similarity respectively.Considering the sparseness of gene-gene matrix,we re-compute the matrix by algorithm KATZ.Then,random walk with restart based on the bilayer network is used to prdict miRNA-target associations.Finally,we carry on fivefold cross validation,recall-precision analysis and pathway enrichment analysis on four networks to prove the performance of our methods.The experimental results show that GLRWR is reasonable and credible.