Entering edit mode
4 days ago
louis-gil
•
0
I've been running the dragen tumor pipelines and I've been noticing discrepancies between the mutations on the vcf output and the BAM file. Both on allele frequency and depth. I'm not sure how to explain this? I would guess that some reads are dropped due to quality. So I see them on the bam but don't make it to the vcf.
You could write to Illumina tech support and ask for resources that explain the default parameters for SNP calling/filtering in DRAGEN pipeline (if you have not been able to find them yourself). Your instinct is probably right and you can check on that by running the DRAGEN BAM through a program/pipeline that you know the contents of/can make changes to. End results may not be identical to what you getting with DRAGEN because of the standard disclaimers of using different programs to the same thing.
I will also say that if you take any variant caller and then look at the corresponding BAM file, you'll find many minor inconsistencies - in the sense that things don't quite add up. Typically, each variant caller will apply an internal heuristic to maximize its performance on the tasks that were important for the authors. These heuristics may be helpful in general, but there are situations when they lead to frustrating behaviors.
I wouldn't worry if the inconsistencies are minor and do not impact the results.
However, some of these inconsistencies may matter a lot, especially when using the tool in a context different from what it was originally designed for.