minus integer copy number when using --purity for cnvkit

Question

[xiaoqiaoliu]

Dear all,

i tried to use estimated purity to adjust the allele specific copy number output from cnvkit, but it would generate minus copy number integer value.

below is my command.

cnvkit.py call sample.cns -v sample.vcf --purity 0.39 -m clonal -o sample_purity.call.cns

below is the output .cns file with minus integer copy number value

chromosome      start   end     gene    log2    baf     cn      cn1     cn2     depth   probes  weight
2       236578628       237490302       AGAP1,GBX2,ASB18,IQCA1,ACKR3    -9.96578        0.00387202      -1      0       -1      26.4653 99      49.9862

did i use cnvkit in wrong way?

thanks.

[2602700485]

Hi, there. I also use estimated purity to adjust integer values. but the cn1 I got -3 and cn2 got 0, what it means? Can I see it as LOH region?

Answer 1

[Kevin Blighe]

It just indicates loss in the sample, i.e., deletion, as evidenced also by the log2 and baf (B allele frequency) values.

The CNVkit manual states that, when cn1 and cn2 differ, it is indicative of allelic imbalance:

Allelic imbalance, including copy-number-neutral loss of heterozygosity (LOH), is then apparent when a segment’s “cn1” and “cn2” fields have different values.

[source: http://cnvkit.readthedocs.io/en/stable/pipeline.html]

There are many copy number programs out there that say that they can detect LOH, allelic imbalance, etc., and none of these has ever made its way to be a clear winner. So, we have many groups using many different programs. Thankfully, one output format has more or less becomes standardised, which is generally referred to as CN segment data:

sample   chr   start   end  num_probes  segment_mean / log2_ratio

You have this data, which is good.

I would not use the cn1 and cn2 values for downstream analyses.

Kevin

[xiaoqiaoliu]

however, how can i get absolute copy number and LOH status for these segment with minus cn?

[Kevin Blighe]

You can output the absolute copy number with the call command.

In this case, it looks like the region is entirely deleted in tumour, i.e., no copies. The BAF is virtually nil / zero, which, coupled with the log2 ratio, implies complete deletion.

log2 of -9.96578 is equivalent to a tumour-to-normal ratio of:

2^(-9.96578) = 0.00100000296

Answer 2

[Eric T.]

You are right that the cn and cn2 values there should be 0, not -1. It's a quirk of rounding -- you can interpret -1 the same as 0, meaning full deletion.

I thought I'd squashed this behavior in recent versions of CNVkit. Are you using the latest version, 0.9.3?

[xiaoqiaoliu]

thanks for the reminding, Eric. i use v0.8.5. i should update to 0.9.3.

[2602700485]

Hi, Eric. I also use estimated purity to adjust integer values. but the cn1 I got -3 and cn2 got 0, what it means? Can I see it as LOH region?