NVIDIA Parabricks v4.6 Enhances Variant Calling Accuracy with DeepVariant and Pangenome Support

NVIDIA has announced the release of Parabricks v4.6, a genomics software suite designed to accelerate secondary analysis for data scientists and bioinformaticians. Published on the NVIDIA Technical Blog on October 14, 2025, the new version offers substantial improvements to variant calling accuracy, leveraging Google's DeepVariant and DeepSomatic 1.9, including a pangenome-aware mode that enhances accuracy across diverse genetic variations and populations. The update also includes enhancements to STAR, a tool for accelerating RNA-sequencing alignment. The key features of Parabricks v4.6 include:

• DeepVariant and DeepSomatic 1.9: Incorporates pangenome-aware DeepVariant for improved accuracy.
• DeepSomatic Support: Adds support for long-read and whole exome sequencing (WES).
• STAR quantMode: Includes GeneCounts for fast generation of gene-level read counts.
• STAR Speedups: Achieves nearly 8x faster performance on two NVIDIA RTX PRO 6000 GPUs compared to CPU-only solutions.
• Mutectcaller Enhancements: Provides additional arguments, including mitochondrial mode.

The integration of Giraffe with pangenome-aware DeepVariant in Parabricks v4.6 significantly improves variant identification accuracy and speeds up the analysis process. Tests using four NVIDIA RTX PRO 6000 GPUs demonstrated a reduction in runtime from over 9 hours on CPU-only solutions to under 40 minutes. STAR also sees improvements, with the new quantMode GeneCounts option enabling fast gene-level read counts during the mapping step, benefiting gene expression analysis, QC, normalization, and data integration. Existing Parabricks users can access DeepVariant by providing the FASTA reference file from the Giraffe index files, a BAM file, and the graph GPZ file output from running Giraffe. Detailed instructions are available in the Parabricks Giraffe documentation. docker run --rm --gpus all --volume $(pwd):/workdir --volume $(pwd):/outputdir \ --workdir /workdir \ nvcr.io/nvidia/clara/clara-parabricks:4.6.0-1 \ pbrun pangenome_aware_deepvariant \ --ref /workdir/hprc-v1.1-mc-grch38.fa \ --pangenome /workdir/hprc-v1.1-mc-grch38.gbz \ --in-bam /workdir/${INPUT_BAM} \ --out-variants /outputdir/${OUTPUT_VCF} docker run --rm --gpus all --volume $(pwd):/workdir --volume $(pwd):/outputdir \ --workdir /workdir \ vcr.io/nvidia/clara/clara-parabricks:4.6.0-1 \ pbrun rna_fq2bam \ --genome-lib-dir ${GENOME_DIR} \ --in-fq ${FASTQ1} ${FASTQ2} \ --output-dir ${OUT_DIR} \ --ref ${GENOME} \ --out-bam ${OUT_BAM} \ --num-gpus ${GPU_NUM} \ --quantMode GeneCounts Users can download NVIDIA Parabricks v4.6 to begin using GPU-accelerated genomic analysis. The NVIDIA Parabricks Developer Forum is available for further discussion and support.