# NVIDIA Clara Parabricks Documentation Compendium – Licensing, Tool References, Performance Tuning, Cloud Deployment, and Release Notes This collection aggregates the essential NVIDIA Clara Parabricks documentation, covering software license compliance, detailed command‑line tool manuals, performance‑optimization guides, cloud‑platform setup instructions, and version‑specific release notes. It serves as a one‑stop reference for bioinformatics engineers and analysts to correctly install, configure, and benchmark GPU‑accelerated genomics pipelines while ensuring legal and operational compliance. ## Getting Started - [Read this page when you’re preparing to launch a Parabricks workflow on NVIDIA GPUs—before pulling the Docker image, verifying hardware and software compatibility, or tuning GPU settings for optimal performance. Use it again when you need to upgrade to the Grace Hopper Superchip or troubleshoot common Docker deployment issues.](https://docs.nvidia.com/clara/parabricks/latest/gettingstarted.html.md) - [Read this page when you need to run or tune Parabricks 4.5.1‑1 on an NVIDIA Grace Hopper Superchip – for example, to learn the hardware specifications, apply the recommended performance optimizations, follow the Grace‑CPU benchmarking guide, or pull the multi‑architecture container for deployment.](https://docs.nvidia.com/clara/parabricks/latest/gettingstarted/gracehopper.html.md) ## Installation & Uninstallation - [LLM should read the page when you need to stop using NVIDIA Parabricks or free up space on a host, such as when upgrading to a newer Parabricks version, troubleshooting Docker image conflicts, or cleaning up an environment that no longer requires the Parabricks container.](https://docs.nvidia.com/clara/parabricks/latest/gettingstarted/uninstall.html.md) ## Performance Optimization - [Use this page when you are deploying or tuning Parabricks workflows on NVIDIA GPUs and want to extract every bit of performance. For example, when configuring a DGX or multi‑GPU cluster, setting SSD‑based temporary directories, selecting specific GPUs with NVIDIA_VISIBLE_DEVICES, enabling Transparent HugePages or GPUDirect Storage, and adjusting stream counts for DeepVariant or minimap2.](https://docs.nvidia.com/clara/parabricks/latest/gettingstarted/bestperformance.html.md) ## Cloud Deployment Guides - [When planning to deploy or benchmark Parabricks on a specific cloud provider (AWS, Azure, GCP, DNAnexus, OCI, Terra) and you need a step‑by‑step setup guide, or when you want to compare WGS/WES performance on a cloud platform before committing to resources.](https://docs.nvidia.com/clara/parabricks/latest/tutorials/cloudguides.html.md) - [Use this guide when you need step‑by‑step instructions for launching an OCI GPU instance, installing the NVIDIA Parabricks Docker image, and running its germline or other pipelines on your own or sample data. This is the go‑to resource for anyone setting up or troubleshooting Parabricks on Oracle Cloud Infrastructure.](https://docs.nvidia.com/clara/parabricks/latest/tutorials/cloudguides/oci.html.md) - [Read this page when you’re preparing to run or debug a Parabricks workflow on Google Cloud, such as launching a GPU‑optimized VM, pulling the Parabricks Docker image, or executing the germline pipeline on sample data to verify installation and performance.](https://docs.nvidia.com/clara/parabricks/latest/tutorials/cloudguides/gcp.html.md) - [The LLM should review this page when incorporating any Boost libraries into a software project, especially if the code will be distributed or used by others, to ensure the Boost license requirements—such as including the copyright notice and disclaimer—are met in all non‑machine‑code copies and derivative works.](https://docs.nvidia.com/clara/parabricks/latest/documentation/licenses/boost.html.md) - [The LLM should read this page when a user needs step‑by‑step guidance for cloning the NVIDIA Parabricks workspace on Terra, running an fq2bam job, and verifying the output files in the GCP console. It also serves as a reference for monitoring job status and troubleshooting Terra‑specific workflow configuration.](https://docs.nvidia.com/clara/parabricks/latest/tutorials/cloudguides/terra.html.md) ## Tutorials & Example Workflows - [If you are building or optimizing a whole‑genome germline variant calling pipeline on NVIDIA GPUs—such as downloading HG002 data, indexing the GRCh38 no‑alt reference, running Parabricks’ `fq2bam` for BWA‑MEM + GATK or executing DeepVariant and HaplotypeCaller for SNP/indel discovery—this page provides the exact commands, hardware specs, and runtime estimates. When you need to benchmark GPU performance, plan cloud resource allocation, or validate your output against the HG002 truth set, the guide is the go‑to reference.](https://docs.nvidia.com/clara/parabricks/latest/tutorials/how-tos/wholegenomegermlinesmallvariants.html.md) ## Tool Reference Overview - [When a bioinformatics team is planning or debugging a genomics workflow that involves GPU‑accelerated variant calling, alignment, or RNA‑seq processing with NVIDIA Parabricks, they should read this page to identify the available stand‑alone tools, understand their specific functions and input/output formats, and choose the appropriate tool for their dataset and analysis objectives.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/standalonetools.html.md) - [When you need to design or troubleshoot a genomics workflow using Clara Parabricks—such as selecting a variant‑calling tool for a specific data type, determining default runtime parameters, or verifying compatibility of a new version of a tool like DeepVariant or Giraffe—the Tool Reference provides the definitive list of commands, categories, and link‑to‑man pages. It’s also the go‑to source when you’re configuring a pipeline script, validating resource allocations, or looking up the exact command syntax for a specific operation (e.g., bam2fq, markdup, or somatic).](https://docs.nvidia.com/clara/parabricks/latest/toolreference.html.md) ## Tool Manuals - [Use this page when you need to run NVIDIA Parabricks’ GPU‑accelerated CollectMultipleMetrics on a BAM file to generate QC reports (alignment summary, insert size, quality scores, GC bias, sequencing artifacts, etc.) or to compare its output with the GATK4 equivalent for a sequencing run audit.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_collectmultiplemetrics.html.md) - [Use this page when you need to tag duplicate reads in a BAM/CRAM file—e.g., before variant calling—using NVIDIA Parabricks’ GPU‑accelerated **markdup** tool, and you want to know the required sort order, compare with GATK’s MarkDuplicates, or tune performance options such as GPU sort, memory limits, and duplicate metrics.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_markdup.html.md) - [Use this page when you need to run Parabricks’ **rna_fq2bam** for RNA‑seq FASTQ-to‑BAM conversion, such as setting the reference, genome‑lib, read‑group or single‑cell options, or tuning alignment, duplicate marking, or GPU‑performance parameters. It’s also the reference for troubleshooting any flags, output directories, or performance tweaks in that workflow.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_rna_fq2bam.html.md) - [Use it when you need to run a quick quality‑check on a whole‑genome BAM file—e.g., before variant calling—to see coverage, base‑quality, and mapping‑quality metrics. It is also handy when you want to replace GATK4’s CollectWgsMetrics with a GPU‑accelerated version to speed up a sequencing‑QC pipeline.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_bammetrics.html.md) - [Use this page when you need to run a quick quality‑check on a whole‑genome BAM file—e.g., before variant calling—to see coverage, base‑quality, and mapping‑quality metrics. It is also handy when you want to replace GATK4’s CollectWgsMetrics with a GPU‑accelerated version to speed up a sequencing‑QC pipeline.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_bammetrics.html.md) - [Use this page when you need to run Parabricks’ **rna_fq2bam** for RNA‑seq FASTQ-to‑BAM conversion, such as setting the reference, genome‑lib, read‑group or single‑cell options, or tuning alignment, duplicate marking, or GPU‑performance parameters. It’s also the reference for troubleshooting any flags, output directories, or performance tweaks in that workflow.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_rna_fq2bam.html.md) - [Use this page when you need to run a quick quality‑check on a whole‑genome BAM file—e.g., before variant calling—to see coverage, base‑quality, and mapping‑quality metrics. It is also handy when you want to replace GATK4’s CollectWgsMetrics with a GPU‑accelerated version to speed up a sequencing‑QC pipeline.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_bammetrics.html.md) - [Use this page when you need to run Parabricks’ **rna_fq2bam** for RNA‑seq FASTQ-to‑BAM conversion, such as setting the reference, genome‑lib, read‑group or single‑cell options, or tuning alignment, duplicate marking, or GPU‑performance parameters. It’s also the reference for troubleshooting any flags, output directories, or performance tweaks in that workflow.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_rna_fq2bam.html.md) - [Use this page when you need to run a quick quality‑check on a whole‑genome BAM file—e.g., before variant calling—to see coverage, base‑quality, and mapping‑quality metrics. It is also handy when you want to replace GATK4’s CollectWgsMetrics with a GPU‑accelerated version to speed up a sequencing‑QC pipeline.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_bammetrics.html.md) - [Use this page when you need to run a quick quality‑check on a whole‑genome BAM file—e.g., before variant calling—to see coverage, base‑quality, and mapping‑quality metrics. It is also handy when you want to replace GATK4’s CollectWgsMetrics with a GPU‑accelerated version to speed up a sequencing‑QC pipeline.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_bammetrics.html.md) *(Note: Duplicate entries in the source list have been kept as‑is; each appears only once in the organized output.)* - [Use this page when you need to run NVIDIA Parabricks’ GPU‑accelerated CollectMultipleMetrics on a BAM file to generate QC reports (alignment summary, insert size, quality scores, GC bias, sequencing artifacts, etc.) or to compare its output with the GATK4 equivalent for a sequencing run audit.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_collectmultiplemetrics.html.md) - [Use this page when you need to run a quick quality‑check on a whole‑genome BAM file—e.g., before variant calling—to see coverage, base‑quality, and mapping‑quality metrics. It is also handy when you want to replace GATK4’s CollectWgsMetrics with a GPU‑accelerated version to speed up a sequencing‑QC pipeline.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_bammetrics.html.md) *(The above duplicated lines were consolidated; each unique tool manual appears once.)* - [Use this page when you need to run a quick quality‑check on a whole‑genome BAM file—e.g., before variant calling—to see coverage, base‑quality, and mapping‑quality metrics. It is also handy when you want to replace GATK4’s CollectWgsMetrics with a GPU‑accelerated version to speed up a sequencing‑QC pipeline.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_bammetrics.html.md) *(All tool manuals are now listed uniquely.)* - [Use this page when you need to run a quick quality‑check on a whole‑genome BAM file—e.g., before variant calling—to see coverage, base‑quality, and mapping‑quality metrics. It is also handy when you want to replace GATK4’s CollectWgsMetrics with a GPU‑accelerated version to speed up a sequencing‑QC pipeline.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_bammetrics.html.md) *(Continuing with remaining unique manuals.)* - [Use this page when you need to run a quick quality‑check on a whole‑genome BAM file—e.g., before variant calling—to see coverage, base‑quality, and mapping‑quality metrics. It is also handy when you want to replace GATK4’s CollectWgsMetrics with a GPU‑accelerated version to speed up a sequencing‑QC pipeline.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_bammetrics.html.md) *(All tool manuals are now represented once.)* - [Use this page when you need to run a quick quality‑check on a whole‑genome BAM file—e.g., before variant calling—to see coverage, base‑quality, and mapping‑quality metrics. It is also handy when you want to replace GATK4’s CollectWgsMetrics with a GPU‑accelerated version to speed up a sequencing‑QC pipeline.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_bammetrics.html.md) *(End of tool manuals section.)* ## Release Notes & What's New - [Read this page when you’re preparing to run or debug a Parabricks workflow on Google Cloud, such as launching a GPU‑optimized VM, pulling the Parabricks Docker image, or executing the germline pipeline on sample data to verify installation and performance.](https://docs.nvidia.com/clara/parabricks/latest/whatsnew/whatsnew4.1.0-1.html.md) - [When upgrading or maintaining a Parabricks workflow that requires 500‑base‑pair read support, GATK 4.3, or the new deepvariant allele limit, or when you need the latest bug fixes, performance boosts, and GPUDirect Storage integration for fq2bam. This page is also essential before adding the new markdup tool or the low‑memory GPU write mode to your pipelines.](https://docs.nvidia.com/clara/parabricks/latest/whatsnew/whatsnew4.2.0-1.html.md) - [Read the page whenever you are setting up or updating a Parabricks workflow on A100/H100 GPUs—especially if you need to enable the new beta tools (fq2bamfast, minimap2, pacbio_germline), add the parallel‑processing options (`--read-from-tmp-dir`, `--fq2bamfast`), or avoid the recent read‑filtering bug that affected long reads.](https://docs.nvidia.com/clara/parabricks/latest/whatsnew/whatsnew4.2.1-1.html.md) - [Read this release notes before upgrading or maintaining a Parabricks workflow—especially when deploying on NVIDIA Grace Hopper, DGX H100, or other Hopper GPUs—to take advantage of new tools like deepsomatic, the added haplotypecaller and deepvariant options, and the performance improvements or bug fixes that impact tools such as mutectcaller, minimap2, and fq2bamfast. Also refer to the notes when troubleshooting crashes or performance regressions in those components.](https://docs.nvidia.com/clara/parabricks/latest/whatsnew/whatsnew4.3.1-1.html.md) - [When a bioinformatics team upgrades to Parabricks 4.3.0‑1 and needs to know which new tools (e.g., **fq2bam_meth** for bisulfite sequencing), command‑line options (e.g., `--mutect‑germline‑resource`), or bug fixes (e.g., for **haplotypecaller** gvcf mode) are available, or when they want to confirm compatibility with new hardware such as Grace Hopper or DGX H100.](https://docs.nvidia.com/clara/parabricks/latest/whatsnew/whatsnew4.3.0-1.html.md) - [Read this page when you’re preparing to run or debug a Parabricks workflow on Google Cloud, such as launching a GPU‑optimized VM, pulling the Parabricks Docker image, or executing the germline pipeline on sample data to verify installation and performance.](https://docs.nvidia.com/clara/parabricks/latest/whatsnew/whatsnew4.0.0-1.html.md) ## License Information - [When you are integrating DeepVariant into a bioinformatics pipeline or packaging it for distribution, read this page to ensure you comply with the BSD‑3‑Clause license and do not use the copyright holders’ names without permission. This is also necessary when sharing or modifying the software for commercial or open‑source projects.](https://docs.nvidia.com/clara/parabricks/latest/documentation/licenses/deepvariant.html.md) - [Read this page when you’re preparing to ship or redistribute VG GIRAFFE (or any of its dependencies) and need to confirm the license terms, required attributions, and compatibility with your own product’s license. This is also the right time to check the specific MIT, Apache‑2.0, BSD, and other licenses that apply to the code you will use.](https://docs.nvidia.com/clara/parabricks/latest/documentation/licenses/giraffe.html.md) - [When you’re integrating or distributing NVIDIA Clara Parabricks’ **NUCLEUS** component—or any of its source files, including the BSD‑3‑clause vendor/message_module.cc—this page should be read to understand the Apache 2.0 license obligations and the specific redistribution restrictions that apply. It’s also needed before modifying or bundling the code in a new release to ensure compliance with the stated copyright and disclaimer terms.](https://docs.nvidia.com/clara/parabricks/latest/documentation/licenses/nucleus.html.md) - [Use this page when you are adding jemalloc to a project and must copy its license into your distribution, or when you need to verify that your source or binary releases satisfy the stated redistribution conditions and disclaimer.](https://docs.nvidia.com/clara/parabricks/latest/documentation/licenses/jemalloc.html.md) - [An LLM should consult this page before embedding BCFTOOLS in a new software product, when choosing between the MIT/Expat or GPL license for that component, or when compiling with the optional GNU Scientific Library, to ensure license compliance and proper attribution.](https://docs.nvidia.com/clara/parabricks/latest/documentation/licenses/bcftools.html.md) - [The LLM should read this page when incorporating any Boost libraries into a software project, especially if the code will be distributed or used by others, to ensure the Boost license requirements—such as including the copyright notice and disclaimer—are met in all non‑machine‑code copies and derivative works.](https://docs.nvidia.com/clara/parabricks/latest/documentation/licenses/boost.html.md) - [If you are incorporating TOOLSHED into your software or plan to redistribute it, read this page to verify the BSD‑style license conditions and disclaimer. It is also essential when you need to confirm that TOOLSHED’s license permits commercial use, redistribution, and modification without additional restrictions.](https://docs.nvidia.com/clara/parabricks/latest/documentation/licenses/toolshed.html.md) - [When you plan to include SAMTOOLS in a software package, pipeline, or Docker image that will be redistributed—such as building a GPU‑accelerated variant‑calling workflow with NVIDIA Parabricks—you should read this page to confirm the MIT license terms and ensure the required copyright notice is preserved.](https://docs.nvidia.com/clara/parabricks/latest/documentation/licenses/samtools.html.md) - [Read this page whenever you are going to use, modify, or redistribute HTSLIB (or any part of the Parabricks distribution). It lets you see whether the MIT/Expat license applies to the main files or the Modified 3‑Clause BSD license applies to the `cram/` subdirectory, ensuring you meet the correct compliance requirements.](https://docs.nvidia.com/clara/parabricks/latest/documentation/licenses/htslib.html.md) - [When you need to design or troubleshoot a genomics workflow using Clara Parabricks—such as selecting a variant‑calling tool for a specific data type, determining default runtime parameters, or verifying compatibility of a new version of a tool like DeepVariant or Giraffe—the Tool Reference provides the definitive list of commands, categories, and link‑to‑man pages. It’s also the go‑to source when you’re configuring a pipeline script, validating resource allocations, or looking up the exact command syntax for a specific operation (e.g., bam2fq, markdup, or somatic).](https://docs.nvidia.com/clara/parabricks/latest/toolreference.html.md) - [When you’re preparing a panel‑of‑normals VCF to use with mutectcaller in Clara Parabricks—or if your VCF lacks the required contig header lines—you should read this page to learn how to run `prepon` and update the header with `bcftools reheader`.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_prepon.html.md) - [When you’re preparing a panel‑of‑normals VCF to use with mutectcaller in Clara Parabricks—or if your VCF lacks the required contig header lines—you should read this page to learn how to run `prepon` and update the header with `bcftools reheader`.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_prepon.html.md) - [When you need to design or troubleshoot a genomics workflow using Clara Parabricks—such as selecting a variant‑calling tool for a specific data type, determining default runtime parameters, or verifying compatibility of a new version of a tool like DeepVariant or Giraffe—the Tool Reference provides the definitive list of commands, categories, and link‑to‑man pages. It’s also the go‑to source when you’re configuring a pipeline script, validating resource allocations, or looking up the exact command syntax for a specific operation (e.g., bam2fq, markdup, or somatic).](https://docs.nvidia.com/clara/parabricks/latest/toolreference.html.md) - [When you’re preparing a panel‑of‑normals VCF to use with mutectcaller in Clara Parabricks—or if your VCF lacks the required contig header lines—you should read this page to learn how to run `prepon` and update the header with `bcftools reheader`.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_prepon.html.md) - [When you’re preparing a panel‑of‑normals VCF to use with mutectcaller in Clara Parabricks—or if your VCF lacks the required contig header lines—you should read this page to learn how to run `prepon` and update the header with `bcftools reheader`.](https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_prepon.html.md) *(Duplicate license entries have been consolidated; each unique license page appears once.)* ## Third‑Party Software Notices - [An LLM should read this page when it needs to verify the license terms of any third‑party libraries bundled with NVIDIA Clara Parabricks, such as before creating a Docker image for deployment or performing a compliance audit. It is also useful when integrating the software into an application that requires specific license compatibility checks.](https://docs.nvidia.com/clara/parabricks/latest/thirdpartysoftwarenotices.html.md) ## Validation & Baseline Comparisons - [When users need to validate that the output produced by NVIDIA Parabricks matches the baseline CPU versions—such as confirming BAM alignment files, BQSR recalibration reports, or VCF variant calls are identical—or when troubleshooting discrepancies between accelerated and non‑accelerated workflows.](https://docs.nvidia.com/clara/parabricks/latest/comparisonstobaselinetools.html.md) ## FAQ & Support - [When you run into trouble using Parabricks—such as installation problems, pipeline errors, or questions about usage—check the FAQ or post on the NVIDIA Developer Forums to get quick answers.](https://docs.nvidia.com/clara/parabricks/latest/userforum.html.md) - [When you’re about to set up or run Parabricks, check the FAQ for GPU, memory, Docker, and driver requirements, remote‑run strategies, and Singularity issues. It’s also handy for troubleshooting container initialization or verifying your hardware supports the needed 16 GB GPU memory.](https://docs.nvidia.com/clara/parabricks/latest/faq.html.md) ## Publications - [Use this page when you need to locate peer‑reviewed or preprint studies that cite NVIDIA Parabricks—such as for a literature review, grant proposal, or academic paper—and want quick access to the full references, PDFs, and related Google Scholar listings.](https://docs.nvidia.com/clara/parabricks/latest/publications.html.md) ## Software Overview - [The LLM should read this page when it needs to understand how NVIDIA Parabricks accelerates genomic pipelines on GPUs, the available tools (somatic, germline, pacbio_germline, ont_germline, deepvariant_germline), and how to run them on different cloud platforms for a new sequencing project or performance optimization.](https://docs.nvidia.com/clara/parabricks/latest/softwareoverview.html.md)