Evaluation of high-throughput isomiR identification tools: illuminating the early isomiRome of Tribolium castaneum

Daniel Amsel, Andreas Vilcinskas, André Billion

MicroRNAs carry out post-transcriptional gene regulation in animals by binding to the 3' untranslated regions of mRNAs, causing their degradation or translational repression. MicroRNAs influence many biological functions, and dysregulation can therefore disrupt development or even cause death. High-throughput sequencing and the mining of animal small RNA data has shown that microRNA genes can yield differentially expressed isoforms, known as isomiRs. Such isoforms are particularly relevant during early development, and the extension or truncation of the 5' end can change the profile of mRNA targets compared to the original mature sequence. We used the publicly available small RNA dataset of the model beetle Tribolium castaneum to create the first comparative isomiRome of early developmental stages in this species. Standard microRNA analysis software does not specifically account for isomiRs. We therefore carried out the first comparative evaluation of the specialized tools isomiRID, isomiR-SEA and miraligner, which can be downloaded for local use and can handle next generation sequencing data. We compared the performance of isomiRID, isomiR-SEA and miraligner using simulated Illumina HiSeq2000 and MiSeq data to test the impact of technical errors. We also created artificial microRNA isoforms to determine the effect of biological variants on the performance of each algorithm. We found that isomiRID achieved the best true positive rate among the three algorithms, but only accounted for one mutation at a time. In contrast, miraligner reported all variations simultaneously but with 78% sensitivity, yielding isomiRs with 3' or 5' deletions. Finally, isomiR-SEA achieved a sensitivity of 25-33% when the seed region was mutated or partly deleted, but was the only tool that could accommodate more than one mismatch. Using the best tool, we performed a complete isomiRome analysis of the early developmental stages of T. castaneum. Our findings will help researchers to select the most suitable isomiR analysis tools for their experiments. We confirmed the dynamic expression of 3' non-template isomiRs and expanded the isomiRome by all known isomiR modifications during the early development of T. castaneum.