Prediction of miRNA in HIV-1 genome and its targets through artificial neural network: a bioinformatics approach

摘要

MicroRNAs (miRNA) are a class of non-coding RNA which inhibits the expression of a particular gene by the process of nucleotide-sequence-specific post-transcriptional gene silencing method. miRNAs are ～21 nt long non-coding RNAs that are derived from larger hairpin RNA precursors. The short length of the miRNA sequences and relatively low conservation of pre-miRNA sequences restrict the conventional sequence-alignment-based methods of finding only relatively close homologs. On the other hand, it has been reported that miRNA genes are more conserved in the secondary structure of their precursor rather than in primary sequences. Therefore, secondary structural features should be fully exploited in the homologue search for new miRNA genes. In this study, an approach for identification and prediction of miRNA in viruses through artificial neural networks (ANN) has been proposed. This idea uses both sequential and structural features of pre-miRNA to train the ANN for identification of miRNA in new viral genomes. The designed ANN was found with an accuracy of 93.68 % for the training dataset and 55.55 % for the validation dataset. In case of HIV, this trained ANN identifies pre-miRNA which does not show sufficient homology to known pre-miRNA sequences, but are highly conserved in their structure. Finally, single miRNA of length 19 mer has been predicted targeting four genes namely NDUFS7, WNT3A, SUFU, and FOXK1 a strict threshold at score 19. The results indicate that this method can be used for identifying novel miRNAs in other viral genomes with considerable success.