small RNA-seq unique reads

Question

[riccardo]

Hi, I would like to know if you use the read counts of reads that map uniquely on the genome or all the reads, unique and not, before to perform a differential expression analysis. In the case of RNA-seq it is better to consider the unique reads, as it is reported in DESeq 2 documentation. It would be great if you had experience not only in miRNAs but also in other small RNAs such as piRNAs. Thank you.

Riccardo

Answer 1

[Asaf]

I think that in the case of miRNAs you should generate a reference of miRNA families and map to this file since you can have the same miRNA or several very similar miRNAs in the genome from the same family. If you'll simply map to the genome and discard non-unique reads you will ignore these miRNAs. I don't have experience with miRNA though, only bacterial sRNAs where the picture is much clearer.

Answer 2

[WouterDeCoster]

For the tools I know about (HTSeq-count, featureCounts), multimapping reads are ignored by default when calculating read counts prior to differential expression.This might be a problem when working on such small RNAs, but ultimately you have to judge whether you have sufficient uniquely mapped reads for your analysis. An interesting suggestion on how to work with multimapping reads is given in this paper: https://genomebiology.biomedcentral.com/articles/10.1186/s13059-015-0734-x in which multimapping reads are aggregated in multi mapping groups for differential expression analysis.

Answer 3

[riccardo]

Thank you!

Riccardo