Calling variants with rastair in the TAPS branch

[davidmasp]

Description of feature

Hi,

My understanding is that rastair should be able to also call/genotype DNA variants, not only methylation. From my understanding, this is not possible with nfcore/methylseq yet right? Would it be of interest?

That is something we would be interested to do with our TAPS data but we would need it to be integrated in some official-ish pipeline.

I have been looking at the current module used in the pipeline. Originally wanted to just modify the $args of the process itself and get the vcf out.¹ I feel with the current module it would be a bit weird to run rastair in calling mode?

Footnotes

1. I also submitted a separate issue in the modules repo regarding a separate problem - here. ↩

[davidmasp]

@eduard-watchmaker fixed the $args issue here: nf-core/modules#12347 👍

From my reading of the docs (https://www.rastair.com/cli.html#rastair-call), it looks like we should be able to generate both VCF and BED outputs by passing something like:

--vcf out.vcf --bed out.bed

Does that sound right?

I'm happy to give this a try, but since these files aren't currently declared as module outputs, it could be a bit awkward to retrieve them from the work directory every time.

I can see two possible approaches for implementing this enhancement:

• Keep rastair call as a single module and allow users to specify optional VCF/BED output filenames via args (with the current stdout behavior remaining the default). We could then expose the VCF and BED files as optional outputs in the module.
• Either within methylseq or nf-core/modules, split the variant-calling functionality into a separate module (e.g. rastair call-vcf or similar). That module would explicitly declare the VCF (and potentially BED) files as outputs.

I'm happy to help with the implementation if that's useful—just looking for some guidance on which direction would be preferred.

[eduard-watchmaker]

Hey David,

rastair2 (which you linked) does indeed call variants (opposed to rastair [1] which is the one in nf-core) but that is still not fully tested nor implemented, but will be in the next couple weeks. We will implement a new feature in nf-core/methylseq to use rastair2 for methylation calling, output a germline variant vcf and use hap.py to benchmark these.

I will definitely keep you in the lop and you can help me develop, implement, test and improve the modules / pipeline!

Let's keep in touch,

Edu

[davidmasp]

Oh okay! completely missed that v1 vs v2 distinction, apologies for that. Great to hear it will be implemented soon, looking fwd to it. Happy to help in any way if needed.

DMP