Clustering with Pyclone using pre-calculated CCFs

[11033...@qq.com]

Hi Andy,

I find that team from the TRACERx project keep using PyClone for clonal deconvolution, but they run PyClone in another way. They calculated CCF manually and just use PyClone for clustering. (e.g. see methods in this paper: https://www.ncbi.nlm.nih.gov/pubmed/29656894). They ignored the CNmajor and CNminor, just use CNtotal, purity and vaf to calculate CCF, and then run PyClone by setting the major_cn=2, minor_cn=0, purity=0.5.

I wonder if you have tried running PyClone like TRACERx in your studies, is there any benefit for doing like this?

Thanks!

Yang

[11033...@qq.com]

Also, as PyClone needs ref_counts and var_counts in input file, I cannot figure out how to run PyClone with pre-calculated CCFs. Could you give some advice? Thanks again!

Yang

[Andrew]

I have not tried running PyClone that way. In essence they are doing the genotype correction first and then just using PyClone to cluster. At that point PyClone is a bit superfluous and any other clustering algorithm could likely be used. I am not sure if that is better or worse.

To your second post, PyClone does not accept CCF input.