mirPD: A pattern-based approach for identifying microRNAs from deep sequencing data

MicroRNAs (miRNAs) are important post-transcriptional regulators of gene expression. In recent years, many novel microRNAs have been discovered at unprecedented depth after advent of deep sequencing technology, but accurate identification of miRNAs is still a challenge due to a large number of isoforms, dynamic range of miRNA expression and unobvious biological characteristics in short mature miRNA sequences.We present a pattern-based approach, mirPD, which uses a two-stage filtration to identify miRNAs from deep sequencing data. In the first filtration stage, patterns capturing conserved knowledge of real miRNAs are extracted from real (published) miRNAs to filter reads. The reads passing the pattern filtration are then mapped to the genome to get candidate precursors which are further filtered according to miRNA biological features in the second stage. Compared with the classic miRNA identification method miRDeep (v1 and v2) on a typical dataset, the experimental result indicates that the mirPD provides higher sensitivity and similar precision, accuracy and specificity.