FFPE Tumor Samples: Poor Goodness of Fit

[Moe94]

Dear Marcus,

once again thank you for providing this tool and the great effort you put into this project.

I am currently struggling to set up PureCN correctly and/or to interpret the results. I have WES data of tumor patients without matched normal. The tumor material is FFPE derived and for quite a lot of these samples I also see quite a lot of noise. Some parameters of my workflow:

• variant calling is done with Mutect2, I use a padded target interval file (50bp), the flags --genotype-germline-sites true --genotype-pon-sites true are used, I use a PoN with 30 samples
• PureCN is then run with the internal segmentation, a bait bed file could not be provided by the manufacturer, so I had to use the target bed file, the NormalDB was created with the PoN generated with Mutect2, there were too few off target counts so I had to remove the --off-target flag and re-run.

For all of my samples I get a poor goodness of fit flag. A log file of representative samples is attached and here is a snippet:

Do you have an idea what might cause this issue? Thank you so much for your help!

Best,
Moritz
S81_copy.log

[lima1]

Another #320 related issue. The log ratio noise is not too bad for FFPE.

Install the latest devel branch and re-run PureCN.R with --min-base-quality 20.