Copy number negative using version 0.9.6

[tgong1]

Hi,

I saw the similar issue #503. As you mentioned it should not present in version 0.9.6, I create this new issue here, thinking it may due to other reasons.

Here how I run CNVKit on a matched tumour and normal samples: ./cnvkit.py batch ${TUMOR_BAM} \ --method wgs \ --normal ${NORMAL_BAM} \ --fasta hs38DH.fasta \ --annotate ${RUN_Dir}/refFlat.txt \ -p 8 \ --output-reference ${RUNDir}/cnvkit${sampleId}.cnn -y --output-dir ${RUN_Dir}

./cnvkit.py call ${RUN_Dir}/${sampleId}-T.final.cns -y -o ${RUN_Dir}/${sampleId}-T.final.call.0.32.cns -m clonal --purity 0.32

Here I show the example results. They have same value of log2, while cn in {0,-1,-2}. chromosome | start | end | gene | log2 | cn | depth | probes | weight chr17 | 39199016 | 39267500 | RPL19,STAC2,LOC101929578,FBXL20 | -9.96578 | 0 | 53.4912 | 78 | 71.7903 chr17 | 39267500 | 39479098 | FBXL20,FBXL20,MIR548BC,MED1,CDK12 | -9.96578 | -1 | 45.1934 | 241 | 221.275 chr17 | 39479098 | 39516852 | CDK12 | -9.96578 | -2 | 28.0059 | 43 | 39.0901 chr17 | 39516852 | 39523876 | CDK12 | -9.96578 | 0 | 50.5651 | 8 | 7.30663

I guess the log2 can be adjusted based on tumour purity, so had a check on the cns file before "cnvkit.py call": chr17 | 39199016 | 39267500 | RPL19,STAC2,LOC101929578,FBXL20 | -0.673364 | 53.4912 | 78 | 71.7903 chr17 | 39267500 | 39479098 | FBXL20,FBXL20,MIR548BC,MED1,CDK12 | -0.886322 | 45.1934 | 241 | 221.275 chr17 | 39479098 | 39516852 | CDK12 | -1.55926 | 28.0059 | 43 | 39.0901 chr17 | 39516852 | 39523876 | CDK12 | -0.710551 | 50.5651 | 8 | 7.30663

Could you give me some idea of the cn = -2 and -1 and how the log2 rescaled based on tumour purity?

Thank you, Tingting

[etal]

Purity of 0.32 is fairly low so the calculation could have overshot. The formula and some examples are shown at the bottom of this section: https://cnvkit.readthedocs.io/en/stable/pipeline.html#calling-methods

I think something different may have happened in #503 where the reported CN value is far below what even naive math would produce.

[tgong1]

Thank you for your reply. I also saw "clonal" method giving me -1 and -2 copy number estimation in other samples with purity around 0.4 and 0.6 and the rescaled log2 values are the same (-9.96578).

As the "clonal" method reporting negative copy number estimation, I'm considering using 'threshold" method. Thank you for referring the section of cnvkit calling method. As "threshold" method estimate copy number from rescaled log2 value, how rescaled log2 value been calculated based on provided tumour purity? In the example I shown, why the rescaled log2 value became the same (-9.96578) while with different original log2 values?

Thank you for your time, Tingting