CNVkit- weird results when using the same sample as normal

Question

[biobiu]

Hi, I'm using CNVkit to detect copy number alterations in hybridized capture panel. When I've used the same sample (exactly the same bam) as a reference I somehow do find some CNA. We used the batch command followed by the call command with default parameters (and with dropping low-coverage). 15 out of the 75 segments were found to have copy-number alteration (cn!=2). The cnn files are exactly the same both for the on- and off-targets so the problem starts when creating the cnr. Any idea why this discrepancy happens?

Answer 1

[Eric T.]

Sorry, I'm not sure I follow. Could you paste both commands here so I can see the difference?

[biobiu]

Thanks for your reply, I'll clarify. There are no two commands, just a batch command of the same sample as test and normal sample:

cnvkit.py batch SAMPLE.bam -n SAMPLE.bam --targets TAR.bed --annotate refFlat.txt --fasta ~ucsc.hg19.fasta --access access-5k-mappable.hg19.bed --output-reference ref.cnn --output-dir --drop-low-coverage

and then: cnvkit.py call SAMPLE.cns

[Eric T.]

This is expected. CNVkit doesn't directly subtract the normal from the test sample; the normal sample is one of several factors that go into the final calls.