Hello all,
Context:
I use the illumina pipeline Dragen for germline analysis on human WGS. For a typical trio i do the germline analysis for each samples to generate a gVCF. Then i use dragen joint genotyping to create a multi sample vcf.
Issue:
Even if for each sample the Callability (% of non N base on a genome with a PASS genotype) is around 97%, when i perform the joint genotyping i get 15 to 20% of variant for which at least one other sample in the trio has a genotype associated to a Bad quality tag (like DragenHardqual). And for exemple if at one position there are 25 reads with the reference and 2 with the variant it put a genotype of 0/1 with a tag DragenHardqual. In this case is obviously a 0/0 in reality, so i set all the genotype with this type of tag to a ./. But, because of that, i can't differentiate the Real unknown genotype (when there are not enough read for exemple) vs the genotype with a Bad genotype that are more probably a 0/0.
as you pay for dragen, you should ask illumina and post the answer here.