RESUMEN
MicroRNAs (miRNAs) are small non-coding RNAs that regulate gene expression at the post-transcriptional level in the cytoplasm and play an important role in a wide range of biological processes. Recent studies have found that the miRNA sequences are presented not only in the cytoplasm, but also in the mitochondria. These miRNAs (the so-called mitomiRs) may be the sequences of nuclear or mitochondrial origin; some of them are involved in regulation of the mitochondrial gene functions, while the role of others is still unknown. The identification of nucleotide signals, which are unique to mitomiRs, may help to determine this role. We formed a dataset that combined the experimentally discovered mitomiRs in human, rat and mouse. To isolate signals that may be responsible for the mitomiRs' functions or for their translocation from or into mitochondria a context analysis was carried out for the sequences. For three species in the group mitomiRs/non-mitomiRs and the group of all miRNAs from the miRBase database statistically overrepresented 8-letter motifs were identified (p-value < 0.01 with Bonferroni correction for multiple comparisons), for these motifs the patterns of the localization in functionally important regions for different types of miRNAs were found. Also, for the group mitomiRs/non-mitomiRs we found the statistically significant features of the miRNA nucleotide context near the Dicer and Drosha cleavage sites (Pearson's χ2 test of independence for the first three positions of the miRNA, p-value < 0.05). The observed nucleotide frequencies may indicate a more homogeneous pri-miRNA cleavage by the Drosha complex during the formation of the 5' end of mitomiRs. The obtained results can help to determine the role of the nucleotide signals in the origin, processing, and functions of the mitomiRs.
RESUMEN
A microRNA (miRNA) is a small noncoding RNA molecule about 22 nucleotides in length. The paper describes a web server for predicting miRNAs and their precursors and binding sites. The predictions are based on either sequence similarity to known miRNAs of 223 organisms or context-structural hidden Markov models. It has been shown that the proposed methods of prediction of human miRNAs and pre-miRNAs outperform the existing ones in accuracy. The average deviation of predicted 5'-ends of human miRNAs from actual positions is 3.13 nt in the case of predicting one pair of complementary miRNAs (miRNA-miRNA* duplex). A useful option for our application is the prediction of an additional miRNA pair. In this mode, the pairs closest to actual miRNA deviate by 1.61 nt on average. The proposed method also shows good performance in predicting mouse miRNAs. Binding sites for miRNAs are predicted by two known approaches based on complementarity and thermodynamic stability of the miRNA-mRNA duplex and on a new approach, which takes into account miRNAs competition for the site. The role of the secondary structure in miRNA processing is considered. The web server is available at http://wwwmgs.bionet.nsc.ru/mgs/programs/rnaanalys/.