Strange CNV level values

[mdozmorov]

Following the Annotating calls and merging over multiple samples tutorial, I created a joint file at 100kb resolution. CNV level values for one region seem to be off-limit:

Exploring the coverage bigWig files didn't show such abnormal coverage.

Exploring the tsv files outputted with cnvpytor -root ${FILEOUT1} -call 100000 > ${FILEOUT11} didn't show this region in the first file, where it shows abnormal CNV level. In the second file, the region looks reasonable (although I still don't understand why CNV level in the TSV is 2.4 and in the joint Excel file is 5).

The issue appears at 100kb resolution and doesn't at 1kb and 10kb resolutions. What may be going on?
Thanks,
Mikhail

[suvakov]

Apologies for the delayed response. The level 2.4331 means that RD here is 2.4 times normal diploid level. It corresponds to CN 5 (in Excel it is rounded to nearest integer but you can adjust to more digits). Level in the first file is strange. Can you share pytor file with me? Thanks

[mdozmorov]

Thanks, Milovan - I shared the file with your .rs e-mail. Please, request permissions if needed.

[suvakov]

Hi Mikhail,

We fixed the bug. It was related to GC correction for bins that are outliers in GC content. As result of this bug one bin (165,100,000) had overestimated RD level (~1e9) and genotyped RD level was artificially high.

You can now update cnvpytor and rerun steps including and after "-his" step for 100kbp bin size.

Thank you for reporting this case and sharing pytor file. It was crucial to reproduce the bug.

Off topic: I see that mean-shift caller failed here to resolve all chr3 events correctly. For cancer samples I recommend you to try our new joint caller.

[mdozmorov]

Thanks, Milovan. Closing as resolved, and I'll try the joint caller.