I am currently trying to scale the usage of my CNVkit implementation to handle a variety of tumor samples that will have unknown purity (anywhere from 10-100%) and was looking for recommendations.
Currently I use the gainloss command (version 0.8.6.dev0) by passing the threshold (-t) 0.585 to detect a single copy number gain. The resulting file is then fed into the call command using no tumor purity estimate.
We have determined that our test set has a mean tumor purity of 55%. I was thinking to use 0.415 as the threshold for the gainloss command before the call command, but the new documentation suggests far lower and I thought perhaps I am misunderstanding the use of the -t flag for gainloss? Can you provide any suggestions for this use case?
Calling Eric T. (developer).