rRNA removal needed for salmon / DESeq2 analysis

Question

[mark.dunning]

Hi all,

I am re-analysing some RNA-seq data that I'm told have high levels of rRNA contamination (not sure how high yet). My usual workflow is salmon -> tximport -> DESeq2

I've seen recommendations that any reads matching rRNA need to removed prior to analysis. Is this still the case with the salmon (alignment-free) protocol? if the rRNA are not in my gtf file, then they will not be quantified and appear in the analysis. Obviously I'll get smaller counts than expected, but should be Ok to proceed with the counts for all the other genes?

Thanks,

[mark.dunning]

So the transcript file I was planning to use from Gencode does include rRNA sequences that I will have to remove before calling salmon

Answer 1

[Devon Ryan]

Unless rRNA is present in your transcriptome fasta file there's no reason to remove it (likewise, there's no reason to remove it in other cases unless it's actually annotated). What you likely should do is quantify how variable rRNA levels are. Sometimes rRNA level ends up being a nice surrogate for other issues (e.g., RNA degradation) with a given sample and you might need to use it for reliably excluding samples.

[mark.dunning]

Thanks! That seems like sensible advice.

[grp2009]

Are there any papers that study these issues (rRNA level as surrogate for RNA degradation, etc.)?

[Devon Ryan]

Probably, search pubmed for them.