RFE: handle circulating tumor DNA (ctDNA) copy number analysis

[lbeltrame]

Since the topic of liquid biopsies is hot right now, and the use of shallow WGS is actually good enough to identify CNVs with approximately 1Mbp resolution, I think it would be nice to incorporate some of the tools I've seen used in the various papers.

Two come to mind:

• QDNAseq from Bioconductor is used quite a bit for ctDNA CNVs, but probably CNVkit would be able to handle that just fine;
• ichorCNA is perhaps the most interesting, because it can also estimate tumor purity (according to the paper's authors, down to 3% tumor fraction). As a downside, it uses a precomputed "panel of normals" and can't be used with other references.

I haven't tested either yet because I don't have any sWGS data at hand. I'd like to, however. Estimates of tumor fraction in ctDNA could also come in handy for downstream post-processing, to actually correct the allelic fraction measured for variants.

Opinions?

[schelhorn]

+1 for ichorCNA, but also because there isn't much else in terms of alternatives.

[chapmanb]

Apologies, misdirected a fix and closed this on accident. Re-opening.

Luca and Sven-Eric -- thanks for this discussion. I'm agreed this would be useful and ichorCNA is my target for doing this as well. We currently have TitanCNA support which I'm working on refining and ichorCNA should follow a similar process. In terms of timelines I have a couple of things I'd like to do before this: do validation of CNVkit calls with/without background samples to model tumor-only calling, look at integration of pureCN for tumor-only cases with this input, then tackle ichorCNA. I'm hopeful general improvements in segmentation and background removal for CNVkit will be useful across all these methods so want to try and see if I can improve that first.

Hope this makes sense in terms of priorities and thank you again for starting this discussion.

[schelhorn]

Sounds great to me, Brad. Thank you.