GenotypeGVCF removes PL field from output

wolfemdwolfemd TheUniverseMember
edited November 2015 in Ask the GATK team

Hi all,

I am trying to combine two VCF files. The first VCF contains data from genotyping-by-sequencing (low coverage) and variants were called with another software (TASSEL). This first VCF has FORMAT fields GT:AD:DP:GQ:PL.

The second VCF was generated by running GenotypeGVCF using the first VCF as the --intervals file. Basically, I am calling the sites from my first VCF out of the whole-genomes contained in the GVCF. I then run CombineVariants on the two VCFs to get the union of both sets of samples.

My problem is that the GenotypeGVCF contains GT:AD:DP:RGQ. There is no more PL field. My guess is that this happened since I used the --includeNonVariantSites option, which was necessary since many sites in my VCF are not variant in the GVCF.

When I CombineVariants I get the union of FORMAT fields. So the samples from the first VCF have PL, those from the GVCF do not. I now want to do imputation (probably with Beagle) and take advantage of the information contained in the genotype likelihoods. Beagle uses the PL field, so the samples without PL would force me to use the GT fields. Using the GT is not okay because Beagle would use that field even for the low coverage data where you might call a homozygote with just one read.

So how can I get PL out of GenotypeGVCF in my situation OR how can I calculate PL, or re-calculate all PL's from the read depth information in the combined VCF?

Thanks in advance!!

Best Answers


  • wolfemdwolfemd TheUniverseMember

    Is there just no answer to this?

    I can be more clear and concise:

    I have a gVCF. I need to extract a list of sites from it to a VCF regardless of whether they are variant or not. I need the PL field present in the gVCF to be preserved. Using --includeNonVariantSites removes the PL field for some reason. Please help!?!?

  • wolfemdwolfemd TheUniverseMember

    Yikes. Ok, thank you. Is there any way to re-calculate PL in a VCF merged from GATK and TASSEL? Using the allele depth information, etc? Or would I have to go back and call SNPs from raw read data?

Sign In or Register to comment.