Are GATK versions 3.* compatible
Dear GATK team,
in exome sequencing analysis, is it in general possible to to indel realignment and BQSR on the bams with e.g. GATK3.1, the HaplotypeCaller with GATK3.2 and the GenotypeGVCFs with GATK3.3?
I ask because I might have mixed different releases and obtained very unexpected results. In general I did the whole exome pipeline (MD,IR,BQSR, HaplotypeCaller, GenotypeGVCFs) first with GATK3.1 and then with GATK3.3, but it is possible that some files were produced by a different GATK version (all greater GATK3 though). I called variants on ca. 30 samples and counted the GT calls of 0/1, 1/1, 0/0, and . and for some samples received very different results. Please take a look at the plot attached. For samples L8333 and L8128 it seems variants were only called with GATK3.3 but not with 3.1. For sample VS0549 it is the other way around. For most samples however (e.g. all the B) samples, the two pipelines are very similar. In general, samples with the same prefix come from the same batch (B, L*, VS* vm*) and have been analyzed together. I will redo all analysis completely with GATK3.3 but it seems to me like this is not the problem because for most samples there is no difference between the versions. The variants have been counted from the raw output of GenotypeGVCFs (no VQSR done).
Have you seen behaviour like this in your data before? From the top of your head do you have any idea what causes this? If not, please don't spent too much time on it sine I am going to redo all sample consitently with the newest GATK version.