Help on understanding CNVkit CN1/CN2

Question

[phusion]

Hi all,

I have recently run CNVkit on some matched tumour-normal samples and I have trouble interpreting the reliability of the CN1 and CN2 estimates. My understanding is that they refer to allele-specific integer copy number values as inferred from total copy number and BAF and that if they differ significantly there is a possibility of allelic imbalance. However, the CN1 values is always lower than the CN2 values for all my samples, which I find highly improbable. I post an example of my output below:

chromosome start end gene log2 probes CN1 CN2 baf weight depth
chr1 10500 34977724 RPL22,MTOR,MTOR,MTOR-AS1,ARID1A -0.191955 254 0.351843 1.39899 0.598086 100.516 720.247
chr1 34977724 80014265 CDKN2C -0.294298 193 38.9794 198.895
chr1 80014265 102870652 RPL5,FAM69A,RPL5 -0.261633 102 21.5914 136.321
chr1 102870652 234590351 NRAS,B4GALT3,PPOX,ELF3,ELF3,RP11-510N19.5 -0.0828552 488 0.381008 1.50737 0.59647 111.538 283.32

Could someone explain to me whether my results are reliable and if not, how should I fix it? For info I've run CNVkit using the bcbio pipeline.

Thanks!

[phusion]

Hi Etal,

Thanks for your reply. Just to let you know that the bcbio pipeline also outputs the results of call when running CNVkit. I checked the columns of the .cns file and the -call.cns file - the BAFs are not the same. I'll also let the bcbio folks know about this issue since I'm not sure whether it's due to the CNVkit script or some processing done by the bcbio folks.

Answer 1

[Eric T.]

Thanks for reporting this problem. I recommend ignoring the CN1 and CN2 values here. If you're using CNVkit directly for follow-up analysis, you can recalculate the allele-specific copy numbers (as integers) using the call command; otherwise, just looking at the baf column will give you roughly the same information where it deviates from 0.5.