I need help with some explanation

Question

[zizigolu]

Hi

For each variant, the variant allele frequency (VAF) depends on the local copy number of the tumor (CPNmut), the purity (p), the local copy number of the normal sample (CPNnorm) and also the cancer cell fraction (CCF), defined as the proportion of cancer cells harboring the mutations. The expected VAF, given the CCF, can be calculated as follows:

VAF (CCF) = p*CCF / CPNnorm (1-p) + p*CPNmut

Then saying

For a given mutation with ‘a’ alternative reads, and a depth of ‘N’, the probability of a given CCF can be estimated using a binomial distribution P(CCF) = binom(a|N, VAF(CCF)). CCF values can then be calculated over a uniform grid of 100 CCF values (0.01,1) and subsequently normalized to obtain a posterior distribution

I got confused here: in the below formula assuming they have CCF, they then calculate VAF based on the CCF

VAF (CCF) = p*CCF / CPNnorm (1-p) + p*CPNmut

Then why again they are going to calculate CCF by saying

CCF values can then be calculated over a uniform grid of 100 CCF values (0.01,1) and subsequently normalized to obtain a posterior distribution

I totally lost the context

Can you help me?

[Ram]

They are calculating VAF(CCF), which I am guessing is a prior probability of the expected VAF for a given CCF.

They then obtain a posterior distribution based on the prior assumption, so the CCF they're calculating is based on the actual observed VAF and not on a theoretical "this should be the VAF for a given CCF" assumption. This needs the first step (prior probability determination) though. Ideally, they should use different notations for CCF_expected/CCF_actual and VAF_expected/VAF_actual

I cannot be completely confident about this answer. If you could point me to the source, I can read up and clarify better.

[zizigolu]

Thank you so much, here

https://stm.sciencemag.org/highwire/filestream/196992/field_highwire_adjunct_files/0/7-283ra54_SM.pdf

In second page, Estimating the cancer cell fraction and mutation copy number section

In my own data I have VAF (alternative allele read counts /total depth) for each variant

I also have tumour copy number

My concerns are:

1- If the actual VAF is my own VAF(alternative allele read counts /total depth)?

2- What is actual CCF?

For this if I am not wrong, I have used CCF function in cDriver R package to calculate CCF (actual ?) using my own VAF (actual VAF?)

3- Which CCF I should use for getting expected VAF?

4- They want expected CCF for getting 95% confidence interval to call a mutation clonal or sub-clonal, so are these all steps required for getting confidence interval? Can not I simply use naive quantile bootstrap 95% confidence intervals (with 10,000 re-sampled averages) on my actual CCF or VAF to call a mutation clonal or sub-clonal?

I have read this document many times but I am just getting more confused

[Ram]

I need some time to read this article. I'll get back to you once I understand what's going on.

[Ram]

I sincerely apologize - I have not been able to pay this sufficient attention. Were you able to figure anything out?

[zizigolu]

Thank you for your time and attention; Actually I am just realising this issue is not that easy and demands much efforts from mysids than a simple communication. I started to know about clonality topic in last recent few weeks and I am just trying to find a software compatible with what input information I can provide

[German.M.Demidov]

I am not sure, but I'd use purity as cancer cell fraction. "Tumor purity is defined as the proportion of cancer cells in the tumor tissue." Absolute does that - I'd try to make it working.