Discussion: consensus CN files for downstream analyses

[jaclyn-taroni]

Related to the generation of copy number consensus calls: #128

Downstream analysis of copy number data, including analysis with GISTIC (#8), often relies on information contained in a SEG file outside of the region, e.g., seg.mean from CNVkit. @fingerfen is generating consensus calls from CNVkit, ControlFreeC, and Manta.

From @hongboxie #128 (comment):

We will present two types of CNVs: Amplification (ploity>2) or Deletion (ploidy<2). We created 3 additional columns where we display original outcome of each method overlapping this region as evidence. The original output of each predictor will carry the information of the ploidy, if it has it.

My interpretation of this comment is that some of the required information for downstream analysis will be retained.

I am filing this ticket to facilitate discussion around how we will use the copy number consensus results for downstream analysis. Because GISTIC is currently run on the CNVkit data (and in my pretty limited experience there seems to be a decent amount of tooling for using CNVkit output), subsetting the CNVkit SEG file to only those regions that have support from the other methods and using that when seg.mean values are required is what comes to mind.

I am also tagging @jashapiro who has been reviewing the consensus code and @jharenza.

[jharenza]

Hi @jaclyn-taroni - I think this sounds like a reasonable approach for CNVkit results/analyses requiring SEG files. Would be happy to rerun GISTIC once we have these consensus calls.

[hongboxie]

@jaclyn-taroni I think it sounds reasonable. Do you have overall plan for downstream analysis of CNV data?

[jaclyn-taroni]

@hongboxie broadly speaking, we'll want to run GISTIC (#8) and plot recurrent alterations in an OncoPrint (#6; though this does not require the segment mean values; also somewhat related to #394). I am assuming we probably want to integrate the CNV data with the SV data for visualization purposes. I've found this UMCCR pipeline code that uses Manta and CNVkit output to create circos plots. I bet we can adapt it (or use it to guide something similar) to use the "consensus filtered" CNVkit data mentioned here, the AnnotSV-annotated Manta file, and circlize.

[hongboxie]

@jaclyn-taroni Sure! Typically circular plot will benefit visualizing genomic translocation. If there is a team working on providing translocation, that will be perfect. Long time ago, we clustered CNVs into groups, to see whether certain tumor or tumor+stages. Not sure if those are something in our plan.

[jaclyn-taroni]

Closed via #441 and the file release is mentioned on #432.