Dear all,
please could you advise, what is a typical range of AF (allele fractions) of somatic mutations in the cancer samples in order to consider them clonal or sub-clonal ? Thank you ...
-- bogdan
To answer your question in a simple sentence - In an ideal scenario, variants with VAF around 50% is fully clonal. Since most somatic mutations tend to be heterozygous, a clonal mutation would have an allele frequency of 50%. Anything below ~45/40% would be a sub-clonal.
But it also depends on your tumor type. Liquid tumors (such as leukaemia) are much clean with little to no contamination. Solid tumors on other hand are dirty, often contamintated with adjacent normal tissues. In that case clonal mutations tend to be clustered around VAF lower than 50%. For example, a tumor with a purity of 80% would have a clonal mutations around 40%.
Things get messier again since one need to consider copy-number status. If a mutation is located on an amplicon, gets its VAF amplified as well. A sub clonal mutation with copy number doubled would have a high allele frequency.
There are algorithms which account for CNV and tumor purity. You can take a look at ABSOLUTE from Broad which estimates tumor purity and clonality for variants. I have seen articles, where ABSOLUTE is used to estimate cell fractions. Here is an excerpt from an article,
Tumor cellularity was determined based on the VAF and segmented copy number data using ABSOLUTE, in order to determine the cancer cell fraction (CCF) of each mutation, as was previously described by McGranahan et al. The clonal status was defined according to the confidence interval (CI) of the CCF. Mutations were classified as subclonal if the upper bound of their 95% CI was less than 1.
But in general, major clone would be around 50%, assuming 100% pure sample.
Again Chris Miller is the right person when it comes to tumor heterogeneity and clonality. He could correct me or add more to it.
Thank you Anand. Happy to be able to use MAFTOOLS (https://bioconductor.org/packages/release/bioc/html/maftools.html) where you have implemented many algorithms for analysis and visualization.
In addition to Kevin's answer, Quantification of subclonal selection in cancer from bulk sequencing data and references therein may be another useful read.
If you want some sensible cutoff for some rough analysis or filtering, I would say 20% AF is about right, but again, as Kevin says it depends...
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
Tagging: Chris Miller