To celebrate the release of GATK 4.0, we are giving away free credits for running the GATK4 Best Practices pipelines in FireCloud, our secure online analysis portal. It’s first come first serve, so sign up now to claim your free credits worth $250. Sponsored by Google Cloud. Learn more at

UnifiedGenotyper false call with no support in BAM

I used GATK 3.2-2 to do indel realignment and multisample calling with UnifiedGenotyper (I cannot use HaplotypeCaller because it is incompatible with the type of data I am analyzing).

Among the ~96 samples sequenced, I have one sample with two nearby variant calls, an indel and a missense, that we checked by Sanger. The missense is real but we found no trace of the indel in the Sanger traces. When I look at indel call in the multisample vcf, it has good allele depth and GQ, but suspiciously has the same AD as the missense call. Additionally, when I look at the bam in IGV, I see no evidence for the indel whatsoever and the variant is not called in any other samples in this project.

chr13:101755523 CA>C GT:AD:DP:GQ:PL 0/1:55,56:111:99:1388,0,1538

chr13:101755530 A>G GT:AD:DP:GQ:PL 0/1:55,56:111:99:2170,0,2125

I went back and recalled just this one sample (single sample calling)… which resulted in the correct variants, i.e. the indel was not called at all, but the SNP, which does validate, is called.

I understand that this is not an easy region to call because of the 7xA repeat, but it’s not obvious to me why this happens only in multisample mode and I'd like to have a better understanding of what is going on.


  • Geraldine_VdAuweraGeraldine_VdAuwera Cambridge, MAMember, Administrator, Broadie

    Hi there,

    Sorry for the late response. It's difficult to comment without seeing the data, but UG does poorly on indel calls in general, so this is probably one of those cases where it is not making the right decision. The fact that this only happens in multisample calling may mean that there is some artifactual evidence for the indel in other samples as well, even though it is not strong enough in the others to generate a call in single-sample mode. But this is only a guess -- like I said, it's hard to say without a lot more information.

  • I could upload a portion of the BAM(s) if someone was willing to take a look.

  • Geraldine_VdAuweraGeraldine_VdAuwera Cambridge, MAMember, Administrator, Broadie

    To be honest we are no longer putting any effort into the UG because it is being replaced entirely by HaplotypeCaller, so we can't take time to look at this in detail.

Sign In or Register to comment.