Finding differentially expressed sRNA-Seq regions with srnadiff

Small RNAs (sRNAs) encompass a great variety of different molecules of different kinds, such as micro RNAs, small interfering RNAs, Piwi-associated RNA, among other. These sRNA have a wide range of activities, which include gene regulation, protection against virus, transposable element silencing, and have been identified as a key actor to study and understand the development of the cell. Small RNA sequencing is thus routinely used to assess the expression of the diversity of sRNAs, usually in the context of differentially expression, where two conditions are compared. Many tools have been presented to detect differentially expressed micro RNAs, because they are well documented, and the associated genes are well defined. However, tools are lacking to detect other types of sRNAs, which are less studied, and have an imprecise \"gene\" structure. We present here a new method, called srnadiff, to find all kinds of differentially expressed sRNAs. To the extent of our knowledge, srnadiff is the first tool that detects differentially expressed sRNAs without the use of external information, such as genomic annotation or reference sequence of sRNAs.\n\nAuthor summaryWe present here a new method for the ab initio discovery of differentially expressed small RNAs. The standard method, sometimes named annotate-then-identify, first finds possible genes, and tests for differential expression. In contrast, our method skips the first step and scans the genome for potential differentially expressed regions (the identify-then-annotate strategy). Since our method is the first one to use the identify-then-annotate strategy on sRNAs, we compared our method against a similar method, developed for long RNAs (derfinder), and to the annotate-then-identify strategy, where the sRNAs have been identified beforehand using a segmentation tool, on three published datasets, and a simulated one. Results show that srnadiff gives much better results than derfinder, and is also better than the annotate-then-identify strategy on many aspects. srnadiff is available as a Bioconductor package, together with a detailed manual: https://bioconductor.org/packages/release/bioc/html/srnadiff.html