Estimate tumor purity for liquid cancer biopsy samples of whole exome sequencing data

Question

[svlachavas]

Dear Community,

briefly, in a current collaboration project, i have performed variant calling analysis on cancer whole exome sequencing data on a small number of 4 patients with small cell lung cancer (Circulating tumor cells sample, biopsy cancer sample and the relative peripheral blood normal sample). Now, except the completed somatic variant calling with GATK4 and mutect2 updated pipeline, I have also performed the suggested GATK2 pipeline for copy number alteration events identification. Currently, my goal is to examine specifically the CTC samples to estimate their relative tumor purity.

The notion for this, is that our collaborators have established a novel biological protocol for the isolation of circulating tumor cells and the subsequent creation of liquid biopsies-thus, an additional estimation of purity (even from the computational perspective) would strengthen the hypothesis if the isolated cells are mainly tumorigenic, or they present contamination with other cell populations (which is expected but the percentage matters..). Thus, is there a tool or pipeline you would suggest for estimating the tumor purity, based on my project description and nature of samples ?

Thank you in advance for your time and consideration on this matter !!

Best,

Efstathios

[German.M.Demidov]

I would give ClinCNV a chance. I can provide guidelines on how to customize it. I've never tried it with liquid biopsy samples and purity <5%, but it still may work after some tweaks.

Answer 1

[amjad]

You can check the ctDNAtools R package if it suits your needs. From a cancer detection perspective, it doesn't matter if your data is from ctDNA or CTC. However, you would need a list of mutations to use for tracking the tumor.

https://github.com/alkodsi/ctDNAtools

[svlachavas]

Dear Amjad,

thank you for your suggestion and idea. Just two relative questions on this matter-your R package is designed only for circulating tumor DNA, and not for biopsies, correct ? Moreover,what particular analyses your pipeline implements, and I would benefit specifically if I analyzed the CTC samples with their normal blood samples ? Finally, you mentioned a list of mutations ? Could you elaborate further on this ?

Best,

Efstathios

[amjad]

There will be a BiorXiv manuscript soon (in 2-3 days) describing what you can do with ctDNAtools. For now you can browse the vignette. Briefly, there are two main analyses you can do: (1) sequencing reads fragmentation: this is only relevant if you have ctDNA data because ctDNA would be fragmented biologically rather than artificially by sonication or other approaches. (2) Minimal residual disease: this can be done regardless of the type of your data. The idea is that you usually can use variant allele frequency (VAF) of somatic variants to get an idea on the the tumor purity (ctDNA concentration in case of plasma). However, in some cases, the tumor purity/ ctDNA amount can be extremely low that you won't have any variants detected. This can happen after good treatment for example. CtDNAtools allows to determine whether this kind of samples have ctDNA or not, if you know what variants the tumor have (from a sample before treatment for example). In your case, if your samples have clear somatic variants, you can use the VAF or copy number alterations to determine tumor purity, and in this case you don't need ctDNAtools.

[svlachavas]

Dear Amjad,

thank you for the additional description-thus far, as i don't have any ctDNA samples, the second scenario would suit. Currently, all the samples (both biopsies and CTCs) are at the time of diagnosis, prior therapy. Thus, i have used tools like FACETS to get an estimation of tumor purity. However, the concept of Minimal residual disease you have mentioned sounds interesting: actually, for one patient, i have also exome sequencing of CTCs after relapse, approximately 4-6 weeks after standard therapy -thus, could I make any comparison using your tool for this patient, regarding the CTC sample at diagnosis versus the CTC after relapse ?

[amjad]

Unfortunately, ctDNAtools wouldn't be useful to compare relapse to primary. But you can check this article for some insight: https://www.nature.com/articles/s41591-019-0561-9.pdf?draft=collection

Hope this helps.