skills/parabricks/references/pbrun-pacbio_germline.md

# Parabricks pacbio_germline

Use this reference for NVIDIA Parabricks `pbrun pacbio_germline` — PacBio germline pipeline from HiFi/CLR FASTQ or BAM/CRAM through minimap2 alignment and DeepVariant calling to VCF/gVCF.

## First Steps

1. Confirm the Parabricks version or container tag.
2. Confirm the data is PacBio germline data and identify HiFi or CLR if
   relevant.
3. Collect required inputs:
   - Reference FASTA.
   - PacBio FASTQ or BAM/CRAM input as supported by the selected version.
   - Output VCF/gVCF or output directory.
   - Model/resource bundle when required.
4. Ask for intervals, sample name, temporary directory, and logs when relevant.
5. For PacBio alignment-only questions, consider
   `pbrun-fq2bam.md` and related FASTQ/BAM references; for runtime readiness, use
   `runtime-environment.md`.

## Command Shape

```bash
docker run --rm --gpus all \
  --volume /host/input:/workdir \
  --volume /host/output:/outputdir \
  --workdir /workdir \
  nvcr.io/nvidia/clara/clara-parabricks:<version> \
  pbrun pacbio_germline \
  --ref /workdir/<reference.fa> \
  <version-specific-input-options> \
  <version-specific-output-options>
```

Verify exact input mode, model, output, and interval flags against the selected
version.

## minimap2/DeepVariant Option Mapping

Use this mapping when translating a baseline minimap2 plus Google DeepVariant
PacBio germline workflow to `pbrun pacbio_germline`. Parabricks v4.7.0
documents this as a long-read pipeline using minimap2 alignment and
DeepVariant calling.

| Baseline option | `pbrun pacbio_germline` equivalent | Notes |
| --- | --- | --- |
| minimap2 reference / DeepVariant `--ref` | `--ref` | Required reference FASTA path. |
| minimap2 index input | `--index` | Optional minimizer index generated by vanilla minimap2. |
| minimap2 FASTQ query input | `--in-fq` | Repeatable FASTQ/FASTQ.GZ input. |
| BAM/CRAM input mode | `--in-bam` | Skip alignment when using prepared BAM/CRAM if supported. |
| minimap2 `-x map-pbmm2` or PacBio preset | `--preset map-pbmm2` or `--preset map-hifi` | Confirm PacBio subtype and selected version before changing the default. |
| minimap2 `-k`, `-uf`, `-ub`, `--MD`, `--eqx`, `-y` | Matching Parabricks minimap2 flags | Parabricks exposes these minimap2 controls directly. |
| GATK/Picard sorted BAM output | `--out-bam` | BAM after sorting/marking where applicable. |
| GATK/Picard `MarkDuplicates -M` | `--out-duplicate-metrics` | Duplicate metrics output. |
| GATK `BaseRecalibrator --known-sites` | `--knownSites` | Repeatable known-sites VCF input when BQSR is used. |
| GATK `BaseRecalibrator --output` | `--out-recal-file` | BQSR report output. |
| DeepVariant `--output_vcf` | `--out-variants` | VCF/gVCF output. |
| DeepVariant `--model_type PACBIO` | PacBio pipeline default, or `--mode`/model flags where documented | Use PacBio-specific mode/model settings from the selected version. |
| DeepVariant custom model | `--pb-model-file`, `--pb-small-model-file` | Parabricks TensorRT model files. |
| DeepVariant `--output_gvcf` / gVCF mode | `--gvcf` plus `--out-variants` | Output path extension controls VCF/gVCF naming. |
| Intervals/regions | `--interval` or `--interval-file` | Parabricks separates inline intervals from interval files. |
| Upstream minimap2/DeepVariant options not listed here | No direct equivalent | Not exposed by current Parabricks docs for `pacbio_germline`. |

## pacbio_germline Options Without minimap2/DeepVariant Equivalents

| `pbrun pacbio_germline` option | Why it has no direct baseline equivalent |
| --- | --- |
| `--pbmm2`, `--pbmm2-unmapped` | Parabricks compatibility controls for pbmm2-style output. |
| `--standalone-bqsr`, `--out-qc-metrics-dir` | Parabricks pipeline mode/output controls. |
| DeepVariant channel/allele-counter flags exposed directly by Parabricks | Parabricks first-class controls for DeepVariant internals. |
| `--nstreams`, `--max-queue-*`, `--chunk-size` | Parabricks minimap2 GPU pipeline controls. |
| `--gpuwrite`, `--gpuwrite-deflate-algo`, `--gpusort`, `--use-gds` | GPU-accelerated write/sort/storage controls. |
| `--low-memory` | Parabricks memory-reduction mode. |
| `--logfile`, `--x3` | Parabricks wrapper logging and full-argument display. |
| `--with-petagene-dir`, `--keep-tmp`, `--no-seccomp-override`, `--preserve-file-symlinks` | Parabricks wrapper filesystem/container controls. |
| `--num-gpus` | Parabricks GPU count. |

## Validation

- Input sequencing technology is PacBio.
- Reference, model/resources, and intervals match.
- Output VCF/gVCF or output directory exists.
- Logs do not show model, long-read preset, reference, mount, CUDA, or memory
  errors.

## Guardrails

- Do not use for ONT or short-read data.
- Do not infer PacBio subtype or model compatibility from filename alone.

## Key References

- <https://docs.nvidia.com/clara/parabricks/latest/documentation/tooldocs/man_pacbio_germline.html>