Support creating a PoN with PureCN

[lbeltrame]

Description of feature

This came up during the Barcelona Hackathon. PureCN can build a panel of normals "database" from coverage data and an optional VCF (https://www.bioconductor.org/packages/devel/bioc/vignettes/PureCN/inst/doc/Quick.html#42_NormalDB).
Since PureCN doesn't support CRAM files at the moment (lima1/PureCN#244) the plan is to use GATK denoised coverage data.

Adding this so I won't forget.

[maxulysse]

@lbeltrame any update on this?

[lbeltrame]

I was about to write about this as I started again to look into this two weeks ago. It is unfortunately a little more complicated than what I had originally imagined. There are two ways to build a PoN with PureCN.

1. Generate (with PureCN) coverage for all the normal samples in the PoN, use a VCF with germline sites genotyped, and create the "normaldb" (this should be covered by the modules in nf-core so they could be used directly)
2. For GATK4 data, it's easier to use the GenomicsDB generated when processing MuTect's PoN (it spares the coverage calculation, because it then relies on GATK4's denoised data). However, this is problematic because PureCN requires the R bindings to the GenomicsDB library, that had never a proper release and exist only as a git repository (last commit ~2y ago). I made a custom Apptainer image for it to test and they work, but I'm worried about the maintainability of this thing.

[maxulysse]

So as far as I can tell, they do releases for PureCN, it's just super hidden.
What we could do is to build a custom container for it, and pin a particular commit.
It will be a pain, but we might have something that could work and be reproducible.

[lbeltrame]

I have a prototype apptainer/singularity file if we want to start from something.