Allele Frequency >0.5 for cancer somatic mutations?

Question

[Rozita]

Hello,

I am analyzing WES VCF files for tumor and matched non-tumor samples and noticed that the allele frequency (AF) values in the files are quite high for the variants (>0.5 and close to 1 for a lot of the variants). This was the case for the variants that I found were present only in the tumor sample but not in the non-tumor sample, so I assume that must mean that the variants/mutations I'm working with are somatic - right? The high allele frequency is what is worrying me a bit - is that normal? Is there another way I could confirm that these variants are somatic. I did not do the variant calling, it was done by the service provider and I'm told that they used GATK for variant calling. I've used SnpEff eff for variant annotation.

Apologies, this is the first time for me to work on this and I don't work with anyone who has expertise around me. Appreciate your help and patience.

Answer 1

[Shred]

The detected AF for a variant is the result of many variables, ranging from tumor stage, sample purity, variant genotype, ploidy and copy number status of the region. It's perfectly acceptable to have frequencies higher than 0.5, and this scenario is way more frequent at later stages of the tumor growth.

Read more here: Systematic pan-cancer analysis of somatic allele frequency [ https://www.nature.com/articles/s41598-018-25462-0 ]

Answer 2

[lethalfang]

It's entirely possible for VAF to be 1 in some cases. To believe it's not some kind of data error, you can look at the tumor-normal alignment in IGV, and make sure it's not low-complexity region, HLA region, very complex alignment (e.g., many indels in the area), low mapping quality scores (good MQ is close to 60), etc.

Answer 3

[Rene]

Somatic variants with a VAF 0.5 - 1.0 are likely due to heterozygous deletion of the remaining allele. In a 100% pure tumour in a diploid sample, you would have a mutation in one allele and loss of the remaining allele, resulting in a VAF close to 1.0. If you include normal-contamination/lower-purity, your VAF will start to drop proportional to the amount of normal contamination, resulting in something like 0.90 for a 90% pure tumour sample. Typically, anything about 0.5 is a result of a copy-number variation on the remaining allele.

[Shred]

In a 100% pure tumour in a diploid sample, you would have a mutation in one allele and loss of the remaining allele, resulting in a VAF close to 1.0

That's true only for variants detected in all the cancer cells. Somatic variants may occur at any stage of tumor growth, resulting in a different clonal distribution.