Identification Of Plant MiRNA Based On SVM And Clonging Of Wheat MiRNA

MicroRNAs(miRNA)ranging in size from 20 to 25 nucleotides represent a new family of noncoding RNAs that function in post-transcriptional regulation of gene expression. They are playing an important role in development of animals and plants as well as resistance against diseases. MiRNAs are processed from long stem-loop hairpin precursors by RNaseâ…¢-like enzyme, Dicer in animals or Dicer-like-l (DCL1) in plants, and then incorporated into a ribonucleoprotein complex, probably identical to the RNA-induced silencing complex (RISC). These tiny RNAs act as small guides and direct negative regulations through sequence complementarity to the 3'-uniranslated regions (UTRs) or coding sequences of target mRNA.In this study, four programs, RP-SVM, GenomicSVM, FS-SVM and mature-SVM, were developed based on novel sequence-structure feature and support vector machine to predict miRNA precursor and mature miRNA. The RP-SVM is a program for identifying real and pseudo pre-miRNA, giving a sensitivity of 91.3.9% and a specificity of 88.4%. The GenomicSVM, an improved method of RP-SVM, gives a sensitivity of 83.3% and a specificity of 98.1%. When applied GenomicSVM method to identify miRNA on genomics or ESTs that contain large number of pre-miRNA-like sequences, it will exclude the false positive pre-miRNA more effectively because of higher specificity than RP-SVM. FS-SVM is a method for predicting functional strand on both stems of pre-miRNA. This program can detect the functional strand on animal dataset efficiently, but it is not efficient when identifying plant miRNA functional strand. So, we developed another program named mature_SVM that was only used to distinguish plant miRNA on pre-miRNA and gave an accuracy of 90% and 76.7% on rice and A.thaliana dataset, respectively.When applied these 4 programs to predict miRNA from the genomics of HBV, rice and A.thaliana, respectively, the results are as following: A) Identifying and validating one miRNA from HBV genomics. This virus miRNA was thought to regulate the gene itself through predicting its target. B) Distinguishing 106 miRNAs in rice genomics and validating 4 out of 10 miRNAs selected randomly from 106 miRNAs, giving an accuracy of 40%. C) Predicting 37 novel miRNAs in A.thaliana genomics. Therefore, we thought these 4 programs were efficient in predicting real miRNA from genomics or ESTs which contains large number of miRNA-like sequences. They provide an alternative method for further identification of novel miRNAs from various organisms.Up to now, miRNAs have been identified in many species including Arabidopsis, rice, C. elegan, mice, and humans. However, there are few reports on the miRNAs in wheat, one of the most important staple crops in the world. The current understanding is that miRNAs are mainly located in the intergenic region or introns of the coding genes. Therefore, predicting miRNAs through computational approaches is limited to model organisms whose genomes have been sequenced. It is hard to predict the miRNAs of other organisms if their genomes have not been sequenced. This discovery not only provides some theoretical basis and experimental evidences for predicting miRNA from wheat ESTs data by using computational biology but also provides an alternative method for predicting miRNAs from ESTs data or genomic sequences.To study wheat miRNA systematically, this paper identified miRNA from wheat though both computational prediction and experiment cloning. Firstly, seventy-nine miRNA candidates were predicted from wheat EST database with developed programs described above. Nine positive signals out of 22 miRNAs that were selected randomly from 79 candidates were detected in special tissue of wheat with northern blotting. On the one hand, small RNAs ( 17~28nt) were isolated and purified by the method of experimental RNAnome, and a cDNA bank was constructed with RT-PCR. Sixty colonies that selected randomly from the cDNA bank were sequenced. Seventeen single sequences were obtained from 41 inserted sequences between 17 nt to 28 nt with homology analysis. Twenty sequences denoted 7 single small RNAs have stem-loop structure through blast searching wheat ESTs and secondary structure predicting with RNAstructure. Moreover, these 7 small RNAs can be detected in special tissues with northern blotting. These 16 small RNAs, including 9 obtained from computational prediction and 7 from experimental cloning , were novel wheat miRNAs according to the principle of biogenesis and expression of miRNA.