Expression profiling of microRNAs by deep sequencing

Abstract

MicroRNAs are short non-coding RNAs that regulate the stability and translation of mRNAs. Profiling experiments, using microarray or deep sequencing technology, have identified microRNAs that are preferentially expressed in certain tissues, specific stages of development, or disease states such as cancer. Deep sequencing utilizes massively parallel sequencing, generating millions of small RNA sequence reads from a given sample. Profiling of microRNAs by deep sequencing measures absolute abundance and allows for the discovery of novel microRNAs that have eluded previous cloning and standard sequencing efforts. Public databases provide in silico predictions of microRNA gene targets by various algorithms. To better determine which of these predictions represent true positives, microRNA expression data can be integrated with gene expression data to identify putative microRNA:mRNA functional pairs. Here we discuss tools and methodologies for the analysis of microRNA expression data from deep sequencing.

Figure 1:

Analytical steps involved with microRNA profiling by deep sequencing. Raw sequences are first filtered to exclude those most likely to represent sequencing errors. Those sequence-reads that align to known microRNA precursor sequences are compiled into a data table with read counts (a measure of absolute expression level). Sequence reads that do not map to known microRNAs may be searched for potential novel microRNAs. In addition to microRNAs, the sequence reads may be searched for other small RNA species, such as piRNAs or snoRNAs.

Figure 2:

Novel microRNA discovery pipeline. Taking the sequence reads that do not map to known microRNA precursors, we proceed by mapping them to the whole genome. Those reads that map (exactly) to the genome are taken (plus 100 bp of genomic sequence flanking either side of the read sequence) and folded as RNA using the Vienna package. The putative novel hairpins this produces are filtered for single-loop hairpins with the putative mature microRNA (read sequence) on one side of the hairpin, and those that pass are kept as possibly valid hairpins. These are trimmed down to only the putative hairpin sequence, refolded and filtered again with the Ambros criteria. The remaining hairpins are a valid microRNA hairpin whose putative mature microRNA sequence was found in the short RNA-ome of the sample, and thus are considered putative novel microRNA precursors.