Why pbrun genotypegvcf only recognizes the last input file when processing multiple GVCFs on a large genome？

Healthcare & Life Sciences Parabricks
1

Hi Parabricks Team,
I am currently using Parabricks v4.6.0-1 to perform joint genotyping on a non-model tree species with a large genome (some scaffolds exceed 1.4 GB). I am encountering an issue where pbrun genotypegvcf only produces a VCF output for the last sample provided in the command line, effectively ignoring all prior samples.

vcf file:

##source=HaplotypeCaller
##GATKCommandLine=<ID=genotypegvcf,CommandLine="pbrun genotypegvcf --ref /workdir/ref/ref.fa --in-gvcf /workdir/R001.g.vcf.gz --in-gvcf /workdir/R002.g.vcf.gz --out-vcf /workdir/2samples.vcf --tmp-dir /hpcfile/users/tas01",Version="4.6.0-1",Date="25/04/2026 22:10:50">
#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  BZ-04
scaffold_1      376     .       G       T       78.60   .       AC=1;AF=0.500;AN=2;BaseQRankSum=-7.030e-01;DP=12;ExcessHet=3.0103;FS=2.932;MLEAC=1;MLEAF=0.500;MQ=47.97;MQRankSum=0.441;QD=6.55;ReadPosRankSum=-4.350e-01;SOR=1.828     GT:AD:DP:GQ:PL  0/1:9,3:12:86:86,0,307

Environment:
Parabricks Version: 4.6.0-1 (Singularity container)
Command:

  • Attempt1:
pbrun genotypegvcf 
  --ref /workdir/ref/ref.fa 
  --in-gvcf /workdir/R001.g.vcf --in-gvcf /workdir/R002.g.vcf 
  --out-vcf /workdir/2samples.vcf
  • Attempt2:
pbrun genotypegvcf 
  --ref /workdir/ref/ref.fa 
  --in-gvcf /workdir/R001.g.vcf.gz --in-gvcf /workdir/R002.g.vcf.gz 
  --out-vcf /workdir/2samples.vcf

[PS: To bypass the TBI file size limit, we compressed the HaplotypeCaller-generated GVCF files using bgzip and then created CSI indices for the g.vcf.gz files using bcftools.]

Technical Observations:

  • I have verified via bcftools query -l that the internal Sample Names (SM tag) for both GVCFs are distinct (R001 and RO02).

  • In the logs, Parabricks only seems to check the index file for the last sample in the argument list:

 \[PB Info 2026-Apr-25 22:10:50\] ------------------------------------------------------------------------------
  \[PB Info 2026-Apr-25 22:10:50\] ||                 Parabricks accelerated Genomics Pipeline                 ||
  \[PB Info 2026-Apr-25 22:10:50\] ||                              Version 4.6.0-1                             ||
  \[PB Info 2026-Apr-25 22:10:50\] ||                               genotypegvcf                               ||
  \[PB Info 2026-Apr-25 22:10:50\] ------------------------------------------------------------------------------
  \[PB Info 2026-Apr-25 22:10:50\] Checking if the index file exists for the input compressed file:/workdir/R002.g.vcf.gz for better performance
  \[PB Warning 2026-Apr-25 22:10:50\]\[src/PBLocalFile.cpp:54\] Could not find index file /workdir/R002.g.vcf.gz.tbi
  \[PB Warning 2026-Apr-25 22:10:50\]\[src/PBLocalFile.cpp:54\] Could not find index file /workdir/R002.g.vcf.tbi
  \[PB Info 2026-Apr-25 22:10:50\] ProgressMeter -  Current-Locus   Elapsed-Minutes
  \[PB Info 2026-Apr-25 22:10:51\] ProgressMeter -  scaffold_1:34049331     0.0
  \[PB Info 2026-Apr-25 22:10:52\] ProgressMeter -  scaffold_1:65603364     0.0
  \[PB Info 2026-Apr-25 22:10:53\] ProgressMeter -  scaffold_1:102366307    0.1
  \[PB Info 2026-Apr-25 22:10:54\] ProgressMeter -  scaffold_1:138099580    0.1
  \[PB Info 2026-Apr-25 22:10:55\] ProgressMeter -  scaffold_1:164822835    0.1
  \[PB Info 2026-Apr-25 22:10:56\] ProgressMeter -  scaffold_1:190346861    0.1

Questions:

  • Why am I unable to successfully perform joint genotyping on multiple (multi-sample) GVCF files?
  • If this is due to an input parameter error, what is the correct syntax for pbrun genotypegvcf to accept multiple input GVCFs in version 4.6.0?

Any guidance would be greatly appreciated, as I need to scale this workflow up to 70 samples.