Automatic curation of large comparative animal MicroRNA datasets

Abstract

Motivation MicroRNAs form an important class of RNA regulators that has been studied extensively. The miRBase and Rfam database provide rich, frequently updated information on both pre-miRNAs and their mature forms. These data sources, however, rely on individual data submission and thus are neither complete nor consistent in their coverage across different miRNA families. Quantitative studies of miRNA evolution therefore are difficult or impossible on this basis. Results We present here a workflow and a corresponding implementation, MIRfix, that automatically curates miRNA datasets by improving alignments of their precursors, the consistency of the annotation of mature miR and miR* sequence, and the phylogenetic coverage. MIRfix produces alignments that are comparable across families and sets the stage for improved homology search as well as quantitative analyses. Availability and implementation MIRfix can be downloaded from https://github.com/Bierinformatik/MIRfix. Supplementary information Supplementary data are available at Bioinformatics online.

The genome of the "Sea Vomit" Didemnum vexillum

“…An additional 20 loci, which harbor two additional families, correspond to previously reported miRNAs [21] which were successfully mapped into the new assembly. To avoid the annotation of false positives due to the modification of the threshold values (see Additional File 1: Figure 12), the position of the mature sequence was evaluated using MIRFix [33] which used both, the RFAM database for the miRNA families alignments and miRBase as source for the annotated mature sequences (as explained in more detail in Methods and Additional File 1: Figure 5). As a result, the definition of a true miRNA candidate relies not only on the homology results given by the sequence/secondary structure comparison, but also in the annotation of their mature sequence.…”

“…In order to annotate the position of mature sequences from miRNA candidates, MIRfix [33] was used. The miRBase (v.22) mature and hairpin sequences were used as initial sequences resource.…”

“…Precursors from predicted miRNAs were obtained in fasta format. The position of the mature sequence (s) from Rfam seed sequences were evaluated using MIRfix [33]. Based on the complete set of corrected seed sequences an evaluation of D. vexillum miRNAs were performed, miRNAs hairpins that contains a candidate mature annotation and are supported by the family structural alignment were considered as true candidates (for details see Additional File 1).…”

miRNAture—Computational Detection of microRNA Candidates

“…The precursor sequences available for a given family often have inconsistently defined ends, and the alignments for different families tend to differ in which species they include. Recent work also identified a moderate number of erroneous entries [19], see also [24].…”

“…On the other hand, utilization of simple, blastbased homology searches alone tend to produce false positives that require extensive curation, which largely relies on the properties expected for a miRNA [17]. The features of canonical miRNAs can be translated in computational rules for the evaluation and editing of structure-annotated alignments of miRNA families that can be utilized to determine whether a candidate sequence fits to a known miRNA family or whether it constitutes a false-positive candidate [18,19].…”