0/0 genotypes in GenotypeGVCFs

Hi,

I want to call both confident variants and confident homREF genotypes using gatk 3.x style (pretty much like EMIT_ALL_CONFIDENT_SITES in UnifiedGenotyper). I first use HaplotypeCaller with -ERC GVCF, then use GenotypeGVCFs -inv to emit all confident sites. However, in the resulting vcf file, there is no site with a 0/0 genotype for all samples in the jointly called vcf, which contains only variant sites and sites with ./. genotype for all samples. I do think all sites are in the vcf file. Why can't I find any site with 0/0 genotype for all samples in the vcf from GenotypeGVCFs? Or do I need to run HaplotypeCaller with -ERC BP_RESOLUTION first? Thanks!

aaron

Answers

  • SheilaSheila Broad InstituteMember, Broadie, Moderator admin

    @aaronchu‌

    Hi Aaron,

    Can you please post a few records?

    Thanks,
    Sheila

  • aaronchuaaronchu USMember

    I attached six gvcf files for which I ran GenotypeGVCFs with -inv to jointly genotype all confident sites. GenotypeGVCFs command:

    java -jar /programs/GenomeAnalysisTK-3.1-1/GenomeAnalysisTK.jar -nt 4 -R hs37d5.fa -T GenotypeGVCFs -inv \
    -V /gvcf/SQC0208F68.hc.gvcf \
    -V /gvcf/SQC0209F68.hc.gvcf \
    -V /gvcf/SQC0210F68.hc.gvcf \
    -V /gvcf/SQC0214F68.hc.gvcf \
    -V /gvcf/SQC0215F68.hc.gvcf \
    -V /gvcf/SQC0216F68.hc.gvcf \
    -D /gatk/gatk_bundle_2.5_b37/dbsnp_137.b37.vcf -o sqcf68.hc.raw.vcf.gz

    I also provided the 0/0 calls from UnifiedGenotyper EMIT_ALL_CONFIDENT_SITES (only SNPs). For all these sites GenotypeGVCFs generates a vcf where there is ./. instead of 0/0.

    Thanks for looking into this!

  • SheilaSheila Broad InstituteMember, Broadie, Moderator admin

    @aaronchu‌

    Hi,

    It turns out this is a limitation of the GATK right now, and we are working to improve it in a future version.

    -Sheila

  • SheilaSheila Broad InstituteMember, Broadie, Moderator admin

    You can also refer here for a similar Forum question. This may help to clarify things: http://gatkforums.broadinstitute.org/discussion/comment/14070#Comment_14070

  • aaronchuaaronchu USMember

    Thanks Sheila! The referred question does have relevance here. It seems the current gatk will not calculate confidence for invariant sites, so I can currently get only confident 0/0 for sites that show at least one variant allele call in the population that is jointly called.
    -aaron

Sign In or Register to comment.