We have some WES data that we are trying to analyze using cnvkit. We identify several interesting regions with cnvkit, but overall the data is fairly noisy. When looking through the .cnr files, I see that much of the noise seems to come from bins that have very low depth in either the reference or the tumor sample. I've tried the --drop-low-coverage option, but the problem seems to persist. Is there a way to manually set a cutoff (we'd like to say at least depth 15 in the control for example).
If there is no effective way to filter out low depths beyond the --drop-low-coverage option, can someone explain a little more deeply how cnvkit handles noisy data caused by poor depth in certain bins and other options that may help?
I've tried manually trimming the .cnr files for low depth which seemed to have no benefit as well as trying to trim the .cnn target files which caused errors presumably because the target and antitarget .cnn files no longer have matching breakpoints.
Thanks for the help in advance.