Over-segmentation issue

[cpreviti]

Hi,
My name is Christopher Previti and I currently work at the German Cancer Research Center (DKFZ). In the clinical project that I work for we use low coverage whole genome data in tumor/normal pairs to determine structural variants. Basically still by eye using coverage comparison plots (ratio tumor-normal coverage). We would like to use WisecondorX to analyse a whole cohort of patient data and I am currently working on specifying the optimal parameters (if possible) for this work. One issue my supervisor has is that she sees a sort of "over-segmentation", where long genomic stretches are split into segments even though there is no apparent reason for this when looking at the coverage plots. I have attached such an example and I was wondering if you would have advise in dealing with such issues. Simply "stitching" these regions together won't work because we would lose a lot of sensitivity...

Thanks for any help you can provide for such a real-world test case.
Christopher
CNV_WiseCondor_hypersegmentation.pdf
.