naumenko-sa
commented
4 years ago Hi @waemm !
Thanks for initiating this discussion!
We have re-enabled PureCN recently (in bcbio 1.2.1), created r36 environment for it, and we definitely need better documentation and validation of PureCN in bcbio to make sure we are running it properly.
My understanding is that currently PureCN receives segmentation information from either cnvkit or gatk-cnv, so one needs to include one of these callers in the yaml to be able to run PureCN.
--rds parameters in the PureCN call is to save PureCN data to make plots, it is not a mapping bias RDS.
PureCN runs on tumor only samples (and it was designed to do so!), but in bcbio in a test run I have it reports that input QC failed:
Cannot find valid purity/ploidy solution. This happens when input
segmentations are garbage, most likely due to a catastrophic sample QC
failure. Re-check standard QC metrics for this sample.
This is most likely a user error due to invalid input data or
parameters (PureCN 1.16.0).
Error: Cannot find valid purity/ploidy solution. This happens when input
segmentations are garbage, most likely due to a catastrophic sample QC
failure. Re-check standard QC metrics for this sample.
This is most likely a user error due to invalid input data or
parameters (PureCN 1.16.0). I think it is because in bcbio we rely on cnvkit or gatk-cnv input for PureCN, and these tools are better suited for T/N case, so there should be a way to better utilize PureCN's potential.
There are definitely better experts on PureCN in bcbio community, please chime in @lbeltrame. @ohofmann maybe you could nominate somebody?
Sergey
lbeltrame
lima1
waemm
roryk
xuwanxing
Hi guys,
I've been trying to run purecn through the bcbio workflow, there is very little documentation on how this works or whether I can provide mapping bias or panel of normal data. Some questions I had:
From looking at the script it looks like a mapping bias RDS is included but how? Is this automatically calculated or do you provide it through the YAML? Any further detail here would be great!
Also, does this run as tumor-only or tumor-normal or can it do both?
Have there been any comparisons of samples run with this workflow to the standard PureCN workflow ?
Thanks!