CNVkit with tumor only bam and flat reference

Question

[youki]

Hi all,

I have a small number of tumor samples without a matched normal control. I would like to know if the following CNVkit batch command, through the generation of the flat reference, is correct:

cnvkit.py batch ${aligDir}/*_Tumor.bam -n -t $WORK/CNV/S07604624_Regions_sorted.target.bed -a $WORK/CNV/S07604624_Regions_sorted.antitarget.bed -f $GENOME/genome.fa \
--access $WORK/tools/cnvkit/data/access-5k-mappable.hg19.bed --drop-low-coverage  -p 20 \
--output-reference my_flat_reference.cnn -d $WORK/CNV_tumor-only \
--diagram --scatter

Thanks

[German.M.Demidov]

I would not recommend since some CNAs are recurrent and they will be "hidden" in your flat reference

[youki]

Yes, I agree @German.M.Demidov. Despite the known limitations when normal controls are not available, I was wondering if this workflow let me compare the bin-level log2 ratios (.cnr) among samples from the same patients.

[German.M.Demidov]

Most likely yes. If the normalization is done in a smart way (I believe it is) then GC-normalized counts will allow you for quite reliable comparisons of levels, if a simple statistical techinque is applied.