estimating tumor purity and ploidy and CNV

Question

[Bogdan]

Dear all,

given a whole-genome-sequencing data, and considering tumo-germline pairs, which algorithm would you recommend for the estimation of purity, ploidy and CNV ?

thank you,

bogdan

[alessandrotestori7]

Can this article be of any help? http://www.nature.com/ng/journal/v47/n4/full/ng.3224.html

Answer 1

[Eric T.]

People have reported good results for your use case with TITAN.

[Bogdan]

Hi Etal, thank you for your comment, and great to hear from you ;) yes, we did set up the pipeline for TITAN indeed, and i was wondering whether in addition to TITAN, shall we consider any other tool;)

[Eric T.]

Well, CNVkit works for WGS too, but estimating purity and ploidy would be a separate step (via THetA2, PyClone, or BubbleTree), whereas TITAN does it all at once. Check the license on TITAN if you're using it for a billed clinical service, though.

Others: CLAMMS is recent and looks good in benchmarks; cn.MOPS is older and does not estimate purity/ploidy, but has always done well in benchmarks and is easy to use.

[Bogdan]

thanks Etal, very useful suggestions ;) on a side note, thought i could ask : have you guys use Sequenza for CNV analysis ?

[Eric T.]

No, but I thought it looked cool based on the paper. It doesn't try to estimate subclone population sizes, so its estimate of tumor purity could be biased lower than that from PyClone or BubbleTree, but it's certainly better than nothing.

[Bogdan]

yes, we will test it. some people were saying that it does better than other tools in terms of finding the focal amplifications.

Answer 2

[poisonAlien]

I have tried many copy number programs (mainly for WXS), but the one which really worked me was Broads yet to relase workflow. It removed so much noice and segmentation was much clear. But again since you have WGS most of the cnv programs should work okay.

You can try ABSOLUTE for purity and ploidy.