The Methods Of Identifying MiRNA

The microRNA(miRNA) is an important type of non-coding RNA,which participates ingene expression regulation.It has been estimated that20~30%of human genes could becontrolled by miRNAs such as the process of cell proliferation,, fat metabolism,differentiation in hemopoietic system etc.Meanwhile,the expression of miRNA and humandisease is closely linked,for instance, the genes relative with cancer and dysnoesia are allinfluenced with miRNA regulation.Many scientists have demonstrated that miRNA genesworked as a core factor in regulation network of the X chromosome vulnerabilitysyndrome.The miRNA studying can give us a depth knowledge of the gene expressionregulation network, and play a very important role for human disease prevention and cure.Tt is estimated that the number of miRNAs in human genome is by the thousand.But thetraditional clone technology can not catch the lower expression miRNAs. Therefore,thescientists began to study calculation prediction methods to discovery new miRNAs ingenome.The purpose of prediction methods is to find out the potential miRNA genes,whichcan be divided into two kinds,one is conservative,the another is not conservative.We use thesequencing data to discovery the miRNA which have been expressed in particular biologicalperiod. In this dissertation,the first part is a summary of the different identitying miRNAmethods. The second part is the recoding process to implement the homolog method and themachine learning method,respectively for identitying conservation and no-conservationmiRNAs in genome.The last part is implementing the particular miRNAs using the highthroughput sequcencing data.The feature of single stem-loops in miRNA is a common filter in the previous homologmethods.With development of miRNA data accumulation,more and more miRNA are presentmulti stem-loops.So,Is the single stem-loops particularly important feature filter foridentitying miRNA or not? The dissertation implement the methods of homology with twoways,one is the filters, another uses two global alignments programinstead of the filter. Atlast,we compare the two methods results and find out that this single stem-loops feature filterdo not a significant effect.I n this dissertation，we verified the influence of the three coding method to the SVMmodel and provided an idea that application of the F-score theory for screening the validfeatures could optimize model and reduce computing time. We refered the principle miRDeep program and analized the sequencing data to discoverythe known and novel miRNAs.