RESUMO
Accurate identification of mature miRNAs is an important requirement for exploring the post-transcriptional regulatory mechanism of organisms. In this work we present a novel computational tool 'Mpred' which first identifies pre-miRNAs and then predicts its mature miRNAs. We first use our method to learn with high accuracy characteristic features of human miRNA precursors from miRbase registry and then apply to sequences from fragile site regions related to cervical cancer in search of novel miRNA genes. The study identified 13 putative miRNA-like sequences and most of them were not related to each other and do not share homology with annotated sequences. Finally, four of the top scoring predictions were verified experimentally using quantitative RT-PCR validation. Expression profile studies revealed that four novel miRs were present in cervical tissues and these data compiled here provide a regulatory framework of miRNA genes that may have roles in tumorigenesis.