Computational Identification And Characterization Of Microrna Genes From Agricultural Insects

MicroRNA (miRNAs) genes, a class of newly discovered-22 nucleotide non-coding RNAs, play important roles in regulating gene expression and organism development by translational repression or cleavage of mRNAs at the post-transcriptional level. In view of their importance, miRNA genes become a research hotspot. MiRNA identification is still an important but difficult task. At present, although thousands of miRNA genes in vertebrates and plants have been identified by experimental or bioinformatics strategies, studies in insect miRNA are focused on several model organisms or genome available pests. Little has been understood about the miRNA genes in agricultural important pests. Here, we established a computational pipeline combined the methods of comparative genomics and machine learning to identify miRNA genes from expressed sequence tags (EST) in 63 insect species. This pipeline employs the Triplet-SVM classifier to recognize real pre-miRNAs from random hairpins. The classifier was trained on the human set of miRNA precursors. In order to achieve good performance in predicting insect miRNA genes, the classifiers were re-trained with known insect miRNA genes. As a result,86 miRNAs from 30 insect species were found based on the strict criteria while 330 miRNAs from 51 species were found based on the loose criteria. The miRNA family analysis indicated that mir-467, mir-297 and mir-466 are much conserved in insect species. We randomly selected nine predicted miRNAs in Locusta migratoria for experimental validation, of which eight miRNAs were successfully verified by RT-PCR. We also predicted 59 insect miRNA target genes, which might also rich in the transcription factors that have the binding and catalytic activities.