Make QC steps opt-in

[edmundmiller]

Summary

Propose making computationally expensive QC steps opt-in across nf-core pipelines via a --run_qc parameter, improving default pipeline performance while maintaining QC capabilities for users who need them.

Champion

@ewels @edmundmiller

Background & Motivation

Many QC steps across nf-core pipelines are computationally expensive and can be the slowest-running parts of the entire workflow. Examples include:

• preseq in methylseq
• Qualimap in both rnaseq and methylseq

Not all users need comprehensive QC for every run, particularly in production environments or when running on large datasets. The current default behavior forces all QC steps to run, regardless of user needs, which significantly impacts runtime and resource usage.

There's a need to establish sane defaults that balance comprehensive quality control with practical performance considerations, while still making it easy for users who need full QC to enable it.

Goals

• Make expensive QC steps opt-in via a --run_qc parameter (or similar mechanism)
• Improve default pipeline runtime and resource efficiency for typical use cases
• Maintain easy access to comprehensive QC for users who need it
• Start with the Very Important Pipelines to establish patterns

Non-Goals

• Removing QC functionality entirely from pipelines
• Breaking existing workflows without providing an easy migration path
• Making QC so hidden that new users don't discover it exists

References

• Original Slack discussion: https://nfcore.slack.com/archives/CP3RJSMF0/p1732017304977949?thread_ts=1731878973.258839&cid=CP3RJSMF0
• Related methylseq discussion on making preseq opt-in

[ewels]

We made a really nice RFC issue template. It's got really lovely fields to fill in, super self explanatory and well documented, easy to use 😉

But yes.. I do think that this is something worth thinking about: reducing QC that may be unused to save resources and help some polar bears.

Of course the devil is in the details. How do we decide which steps are crucial and not, and should be on / off? Does the compute time for a step affect whether it always runs? These are the kinds of things that this RFC needs to solve to make it into a new Guideline Recommendation.

[edmundmiller]

Sorry, I just wanted to convert this from a draft issue that I had floating around because, apparently, you can't Google draft issues.

[jfy133]

This is so incredibly vague I'm tempted to just close it 😅 but will keep it open for now for you to update @edmundmiller ;)

But it seems also extremely pipeline specific that it does not warrant an RFC.

Qualimap for example will only effect a few pipelines.
The importance of QC metrics will depend on a case by case basis.

Unless the proposal is that guideline should be 'defaults tool execution should be as streamlined as possible'? But that is not my impression here.

[jfy133]

@edmundmiller do you still plan to re-make this into an offiical RFC? Otherwise I would close this

[mashehu]

@edmundmiller one more ping a month later 😁