miRTRAP, a computational method for the systematic identification of miRNAs from high throughput sequencing data

microRNAs (miRNAs/miRs) are small regulatory RNAs present throughout the Eukarya [1-3]. They modulate diverse biological processes, including embryonic development, tissue differentiation, and tumorigenesis. miRs inhibit translation and promote mRNA degradation via sequence-specific binding to the 3' UTR regions of mRNAs [2]. They are produced from hairpin precursors (pri-miRNAs) that are sequentially processed by Drosha/DGCR8 and Dicer to generate one or more 19- to 23-nucleotide RNAs. The most abundant product is referred to as miR, while the less abundant sequence produced from the opposite arm of the hairpin is called miR*. In addition, it has been observed that some miRNA loci can produce up to two additional products immediately adjacent to the miR and miR* sequences, which are called miRNA offset RNAs (moRs) [4,5].The comprehensive identification of the complete set of miRs is complicated by their small size, which limits simple cross-species comparisons based on sequence homology. Moreover, de novo computational miRNA prediction methods rely heavily on known miRNAs and are not always effective for characterizing novel genomes. Recent advances in high throughput sequencing technology provide an opportunity for the systematic identification of every miRNA gene in a genome. Here we present such a system for the computational identification of miRNA genes from deep sequencing data and apply it to datasets collected from different developmental stages of the simple chordate Ciona intestinalis. This approach predicted over 300 novel Ciona miRNAs and revealed the molecular phylogeny of miRNA families in the chordate lineage. This method was also used to identify novel miR loci in the extensively characterized genome of Drosophila melanogaster.The comprehensive identification of the full repertoire of miRNAs in a given organism is of general interest. Early bioinformatics approaches used machine learning and pattern recognition to predict miRNA loci de novo from who