Implicit Assumption of Germline Short Variant Discovery
Is an implicit assumption of the pipeline that the normal sample is far away enough from the tumour sample, such as a blood sample, so that cancer cells are not present at low levels? I have been seen a couple of variants with allele frequency of about 20% appearing in the germline variant file and the same variant has about the same frequency in the matched adjacent normal cancer sample taken less than one centimetre away from the visible tumour. I think this should be made explicit in the documentation and perhaps also discussing Broad Institute's DeTiN.
-
Hi Dario
You are certainly right that there may be tumor in normal contamination present in normal samples, however that may not be the only reason why you are observing such variants.
1- Limited somatic mosaicism could be a reason for observing 20% alt allele fractions in normal samples that could have made into to tumor as well (This is something I personally observed in some germline cases as well).
2- Contamination during sample preparation and sequencing could well be another reason why we have Contamination Estimation workflow within our FilterMutectCalls tool.
Contamination cannot be solely due to a single reason therefore we have means to calculate rough tumor segmentation before estimating whether a particular variant is germline or somatic in nature. DeTiN also indicates that such reasons could very well offshoot estimates for tumor in normal contamination therefore even if you use that method one should be extremely careful in interpretting the results.
Most somatic workflows require additiona post processing due to other needs therefore we do recommend users to follow our guide first and then perform any additional necessary post processing steps they see fit.
I hope this helps.
Regards.
Please sign in to leave a comment.
1 comment