During AnnotateCohort we read in the MatrixTable of variant data, and the VQSR-VCF of QC'd data. We hop the INFO section of the VCF over to the MT, containing new filters to be applied. We don't shift the new VQSR header rows over to the MT.
This means that when we make a VCF representation from the MatrixTable, the VCF contains the filters but not the required header lines for downstream tools to parse them.
During AnnotateCohort when we combine VQSR and MT, also add the new header lines over to the MT, so any future VCF representations will be spec-compliant
This has caused some failures to parse VCFs downstream, but these failures are non-blocking as we have workarounds
During AnnotateCohort we read in the MatrixTable of variant data, and the VQSR-VCF of QC'd data. We hop the INFO section of the VCF over to the MT, containing new filters to be applied. We don't shift the new VQSR header rows over to the MT.
This means that when we make a VCF representation from the MatrixTable, the VCF contains the filters but not the required header lines for downstream tools to parse them.
During AnnotateCohort when we combine VQSR and MT, also add the new header lines over to the MT, so any future VCF representations will be spec-compliant
This has caused some failures to parse VCFs downstream, but these failures are non-blocking as we have workarounds