Small non-coding RNAs, often abbreviated as sncRNAs, are short RNA molecules that play important roles in regulating gene expression. Among them, microRNAs, or miRNAs, are especially well known for controlling how messenger RNA is translated into proteins. Because of their role in many biological processes and diseases, including cancer, accurately identifying and measuring these molecules is a key goal in modern RNA sequencing workflows.

One challenge in small RNA sequencing is correctly classifying and counting these short RNA molecules. Their small size and frequent modifications make them harder to analyze compared to longer RNA types. To address this, researchers at the University of Cambridge developed a new computational tool called PymiRa.

PymiRa is designed to quickly and accurately identify miRNAs from sequencing data files. It uses a combination of computational approaches, including a Burrows Wheeler alignment strategy, to match sequencing reads to known miRNA reference sequences. Importantly, the tool allows for small mismatches at the end of reads, which helps account for natural variations and modifications that often occur in miRNAs.

Normalised miRNA counts to the Real (actual) count, showing the differences from each aligner (PymiRa, Chimira, miRDeep2, and Bowtie2) on A) the whole simulated miRNA dataset, and B) the top 600 over-expressed miRNAs from the simulated miRNA dataset.

The simulated 10 million read dataset was comprised of three million real miRNA sequences and seven million shuffled control sequences. The red horizontal line at y = 0 indicates zero deviation from the Real counts. Deviations from the red line indicate a difference in each aligner’s count for each individual specific miRNA, compared with the respective count obtained from the Real simulated miRNA dataset. Of the real miRNA sequences, 600 of these were randomly selected to be given increased counts to simulate a disease model where there is miRNA dysregulation (i.e., over-expression of specific miRNAs). Pearson’s r scores were calculated for each aligner comparing all the aligner’s counts with the Real counts. Bowtie2 was run with different settings/conditions, combined with either HTSeq or FeatureCounts (FC). The settings used were standard default (std), –very-sensitive-local (vsl) and –local -N1 (local). Bowtie2 (std_FC) was used to represent Bowtie2 for evaluating performance for the top 600 miRNAs due to superior performance to other conditions. The average Real count per miRNA was ~ 1,140. The miR-548 family of miRNAs is highlighted, as all aligners struggled to generate accurate counts for these miRNAs, as expected and as described in the text, due to sequence homology

Unlike earlier tools that rely on a single alignment method, PymiRa combines different strategies to improve both accuracy and efficiency. This makes it particularly useful for large datasets generated by RNA sequencing experiments. By better handling post transcriptional modifications, the tool can provide more reliable counts of miRNA expression levels.

The availability of PymiRa as an open resource means that researchers can easily incorporate it into their workflows. This will help scientists better understand how small non-coding RNAs function in normal biology and how their dysregulation contributes to disease. Improved analysis tools like this are essential for advancing research in gene regulation and developing new diagnostic or therapeutic strategies.

Availability – The code for the PymiRa tool is available on the GitHub page https://github.com/ZScurlock/PymiRa