I am merging few hundred of samples for a project level VCF. The following summarize my steps:
a) performed a combineGVCF on a set of gVCF (pVCF1) and then a combineGVCF on another set of gVCF (pVCF2)
b) performed the genotypeGVCF on pVCF1 and pVCF2
c) ran VQSR on this genotypeGVCF output.
What I found is there are variants found in output of genotypeGVCF, but not in pVCF1 and pVCF2, and they all pass the variant filters (VQSRTrancheSNP99.80to99.90 or VQSRTrancheSNP99.70to99.80). I am confused why I am getting these results.