Unsupervised Learning of Multiple Motifs in Biopolymers Using Expectation Maximization
Machine Learning - Special issue on applications in molecular biology
| Hi-index | 0.00 |
MicroRNAs (miRNAs) are non-coding RNA molecules that bind to and translationally repress mRNA transcripts. Currently ~1345 miRNAs have been identified in at least twelve species through experimental and computational approaches. Here, we report on a field not yet thoroughly investigated: the transcriptional regulation of miRNAs. Adequately locating miRNA promoter regions will provide a reasonable search space for computational and experimental studies to determine regulatory factors that drive miRNA transcription. Insight in to the factors that control miRNA transcription may provide clues regarding more complicated mechanisms of miRNA expression control in a developing organism. We use a novel Expressed Sequence Tag (EST) based approach to approximate promoter regions for intergenic miRNAs in order to detect specific and over-represented regulatory elements. We find that miRNA promoter regions may be enriched for binding sites that recruit transcription factors (TFs) involved in development, including several homeobox TFs such as HOXA3 and Ncx. Additionally, we use clustering techniques to cluster miRNAs according to tissue specificity to find tissue-specific regulatory elements. We find a few over-represented binding sites in brain-specific miRNA promoter regions, some of which recruit TFs involved specifically with the development of the nervous system. Based on the results we suggest an interesting mechanism for in vivo miRNA expression control. The EST-based pri-miRNA assembly program will be made available at the website of the DIANA-group by the time of publication (http://diana.pcbi.upenn.edu).