tumor fraction estimation for ctDNA panel sequecing data

Question

[J.F.Jiang]

Hi all,

Tools such as PureCN can estimate purity for tumor samples based on CNVKit outputs, I found it hard to estimate the purity for ctDNA samples.

Since ichorCNA can only be applied for WGS data, I am wondering if there is any tool that can do this prediction (besides ichorCNA_offtarget)?

Thanks, Junfeng

Answer 1

[markus.riester]

Not sure that is very helpful, but we are using a relatively simple XgBoost model that extracts features from PureCN output. Things like autocorrelation and range of the log2 ratios, allelic imbalances of SNPs, 90 percentiles of allelic fractions of somatic mutations, hotspot mutations, etc. I also add a couple of Picard QC metrics so that the model learns to distinguish noise from signal.

[J.F.Jiang]

great, is it open public for share or just for self-usage? Thanks

[markus.riester]

Unfortunately not available, it’s tailored to our pipelines. I might add a section to the PureCN tutorial, but no timeline yet. But it’s pretty straightforward getting those basic features I listed out of the output.