Coverage command for FFPE tissue

[shashwatsahay]

Hi @etal

Just wanted to know if cnvkit somehow corrects for double counting from overlapping paired end reads caused in ffpe tissue....

As far as i could see, CNVkit internally uses pysam.bedcov which I think doesnt have pair awareness. I know this would be hopefully be corrected if we had a ffpe normal reference panel but as you know in clinical world normal's are hard to come by and I sometimes need to run the pipeline using a flat reference.

Do you think it is feasible or generally a good idea to use samtools depth with -s for this. I would be happy create a pull request if you think this is a good idea and won't break any assumptions that cnvKit makes

[etal]

Great point. Could you open a PR for this? And, do you have a way to compare its results to the status quo, at least to get a sense of the magnitude of difference it makes?

[shashwatsahay]

I’m wrapping up testing on the overlapping reads fix using a large FFPE cohort and expect to open a PR by the end of the week. I’m validating this by checking if artifactual focal copy numbers are successfully removed (verified via IGV) when overlapping reads are handled correctly.

Also, are you open to a PR for a PSCBS segmentation workflow? I’ve implemented a version that uses both LogR and DH (similar to PureCN’s approach but tweaked for VCF inputs). In my testing, this joint segmentation detects breakpoints in FFPE samples much better than the standard LogR-only CBS method. It also fully supports tumor-only mode.

[etal]

Joint segmentation PR: yes, definitely! That would be a great addition.