CombineVariants fail when combining vcf from genotypeGVCFs genotyped using -allSites flag

When I attempt to run the tools CombineVariant with two vcf files generated from GenotypeGVCFs using the flag -allSites I get an error saying that I should not use combineVariants in a gvcf file. But my files are not gvcf.

Error:

ERROR MESSAGE: CombineVariants should not be used to merge gVCFs produced by the HaplotypeCaller; use CombineGVCFs instead

Command used:
java -jar GenomeAnalysisTK.jar
-T CombineVariants
-R /path_to_reference/REFERENCE.fa
--variant /path_to_genotypeGVCFs_output/species1_allsites.vcf
--variant /path_to_genotypeGVCFs_output/species2_allsites.vcf
-o allcombined.vcf
-genotypeMergeOptions UNIQUIFY

gatk version v3.7-0

If I run it in the vcfs from GenotypeGVCFs not using the flag -allSites, I don't get this error and it runs fine.

Issue · Github
by Sheila

Issue Number
2425
State
closed
Last Updated
Assignee
Array
Milestone
Array
Closed By
vdauwera

Answers

  • SheilaSheila Broad InstituteMember, Broadie, Moderator

    @luanalins
    Hi,

    Hmm. Let me check with the team whether this is expected. I will get back to you soon.

    -Sheila

  • Geraldine_VdAuweraGeraldine_VdAuwera Cambridge, MAMember, Administrator, Broadie

    Ah, no this is not really intentional. This is due to some overzealous checks we put in to make sure people wouldn't use this tool on GVCFs. The logic of the checks was a bit too naive as we didn't foresee this use case. Unfortunately it's not something we can fix anymore as we've stopped development on 3.x versions.

    If you're trying to compare members of different species, I would recommend just joint-genotyping all of them together rather than comparing calls post-genotyping.

Sign In or Register to comment.