rnaseq
rnaseq copied to clipboard
nf-core
Clarify infer strandedness from current subsampling + infer step
Description of feature
The current nf-core/rnaseq (3.12.0) has initial steps to infer strandedness by first subsample fq, then Salmon to infer strandedness. This is an optional step and has led to some confusion as it's not actually subsampling all the reads.
On a nf-core slack discussion, @drpatelh suggested:
Maybe subsample + infer need to be part of the same station. I think we chose to do it this way because it would have meant introducing more lines and curves to the map which would make it even more confusing. Can you create an issue for this please.
The suggestion is to combine into one station / one module or workflow step. That would clean up the metro diagram and avoid the confusion.
This could be called
- "Infer strandedness (fq, Salmon)" in the metro diagram
- "Auto-infer strandedness by subsampling and pseudoalignment (fq, Salmon)" in the list of steps.
To even complicate matters, the most recent release of BBTools (39.03) now also contains a new tool to infer strandedness called checkstrand.sh.
I have not done any comprehensive evaluation, but it has a samplerate=1.0 parameter and can also stop preliminarily after a fixed number of reads reads=-1. Since it is a one-stop shop written by a reputable author, I believe, chances are that it is way faster than the current subworkflow?
@ewallace - does https://github.com/nf-core/rnaseq/pull/1307 fix things for you?
The new subway map is labeled (Salmon, fq) - I agree that the Salmon is more important than the fq, but fq happens before subsampling, so you may wish to switch the order in which they are written on the subway map depending on your goals.
yeah, I saw the comment and modified my PR in accordance, and then said I've done it
yeah, I saw the comment and modified my PR in accordance, and then said I've done it
You were too fast for me, I didn't think you'd already have addressed the comment. All good now :-)