Accuracy of microRNA discovery pipelines in non-model organisms using closely related species genomes

Abstract

Mapping small reads to genome reference is an essential and more common approach to identify microRNAs (miRNAs) in an organism. Using closely related species genomes as proxy references can facilitate miRNA expression studies in non-model species that their genomes are not available. However, the level of error this introduces is mostly unknown, as this is the result of evolutionary distance between the proxy reference and the species of interest. To evaluate the accuracy of miRNA discovery pipelines in non-model organisms, small RNA library data from a mosquito, Aedes aegypti, were mapped to three well annotated insect genomes as proxy references using miRanalyzer with two strict and loose mapping criteria. In addition, another web-based miRNA discovery pipeline (DSAP) was used as a control for program performance. Using miRanalyzer, more than 80% reduction was observed in the number of mapped reads using strict criterion when proxy genome references were used; however, only 20% reduction was recorded for mapped reads to other species known mature miRNA datasets. Except a few changes in ranking, mapping criteria did not make any significant differences in the profile of the most abundant miRNAs in A. aegypti when its original or a proxy genome was used as reference. However, more variation was observed in miRNA ranking profile when DSAP was used as analysing tool. Overall, the results also suggested that using a proxy reference did not change the most abundant miRNAs' differential expression profiles when infected or non-infected libraries were compared. However, usage of a proxy reference could provide about 67% of the original outcome from more extremely up- or down-regulated miRNA profiles. Although using closely related species genome incurred some losses in the number of miRNAs, the most abundant miRNAs along with their differential expression profile would be acceptable based on the sensitivity level of each project.

Figure 1. Number of reads from A. aegypti smRNA-seq data mapped to four selected species unique mature miRNA sequences in miRBase.

Figure 2. Number of reads from A. aegypti smRNA-seq data mapped to four selected species sequence of pre-miRNA hairpins in miRBase.

Figure 3. Number of conserved or known miRNAs identified in A. aegypti smRNA-seq data by miRanalyzer and DSAP.

A) The total number of identified miRNAs with strict and loose criteria by mapping to species unique sequences in miRbase. B) Number of overlap or common miRNAs between A. aegypti and other selected species.

Figure 4. Percentage of unique reads from A. aegypti smRNA-seq data aligned to genome sequences from four different insects using strict or loose criterion.

Figure 5. Number of novel miRNA candidates predicted in A. aegypti smRNA-seq data by mapping to four selected species genomes.

“Perfect Dicer pattern: A perfect 3′ 2 nt overhang exists for the most expressed read and not more than 3 read clusters do exist on the pre-miRNA (one for the mature, one for the mature* and one for the loop sequences). Diffuse Dicer patter: A 3′ 1–4 nt overhang exists for the most expressed read and not more than 3 read clusters do exist on the pre-miRNA. Low fluctuation: No Dicer pattern has been detected but only one read cluster does exist. No Dicer pattern: No Dicer pattern has been detected or too many read cluster do exist” .

Figure 6. Unique highly differentially expressed miRNAs in Aag2 cell with and without Wolbachia infection when genomes of different selected insect species were used as references.

Overlap areas show the number of common miRNAs in each comparison with high level of fold changes (more than 2).