RNA stem-loops: to be or not to be cleaved by RNAse III

Most of the vertebrate genome is transcribed into RNA. Transcribed regions contain hundreds of thousands of potential duplex structures that could serve as substrates for RNAse III enzymes of microRNA (miRNA) maturation pathways. Yet, only a minority of these potential precursors make their way to the cytoplasm to form mature miRNAs. We question here what specific structural features make an RNA stem-loop structure an adequate primary or precursor miRNA. We address this question by comparing known pre-miRNAs to other predicted noncoding transcripts obtained from comparative genomics scans, using the structure comparison program RNAforester. By analyzing a classification tree of 1200 such RNA structures, we observe that pre-miRNAs cluster distinctly from other duplex structures of apparently similar size and free energy. The most distinctive features of nonprecursor duplexes are increased lengths and numbers of bulges and internal loops when compared to real miRNA precursors. Thanks to these characteristics, secondary structure comparison can predict the miRNA precursor status of a candidate stem-loop with a surprising accuracy. Furthermore, predicted noncoding transcripts tend to depart from miRNA precursor characteristics more strongly than randomly occurring duplex structures in genomic DNA. This result suggests that many noncoding RNAs may be under selection to dodge the RNAi pathway.