A Nextflow DSL2 pipeline for GPU-accelerated basecalling of Oxford Nanopore POD5 files using Dorado. Designed as an upstream workflow that generates basecalled BAM and FASTQ files for use in downstream analyses.

This workflow:

1. Downloads the specified Dorado basecalling model
2. Builds a minimap2 index of the reference genome
3. Basecalls each POD5 file in parallel (GPU-accelerated)
4. Merges per-POD5 BAM files by sample
5. Converts merged BAMs to compressed FASTQ

• Nextflow ≥ 22.10.0
• Singularity (container runtime; Docker not supported by default profiles)

All bioinformatics tools (Dorado, minimap2, samtools, pigz) are provided via the container oras://ghcr.io/shians/dorado-container:1.1.1-singularity.

An NVIDIA GPU (A30 or compatible) is required for basecalling. High-performance storage is recommended for POD5 files (1–5 GB per file).

A tab-separated file mapping sample identifiers to directories containing POD5 files.

Format (pod5_sheet.tsv):

sample_id	path
control_rep1	/data/nanopore/pod5/control_replicate1
control_rep2	/data/nanopore/pod5/control_replicate2
case_rep1	/data/nanopore/pod5/case_replicate1_flowcell1
case_rep1	/data/nanopore/pod5/case_replicate1_flowcell2

• sample_id: Identifier for each sample (used to name output files). Multiple rows with the same sample_id are merged into a single output file, which is useful when one sample was sequenced across multiple flow cells or run folders.
• path: Absolute path to a directory containing .pod5 files (searched recursively)

A FASTA file used for alignment during basecalling (via minimap2). Gzipped FASTA (.fa.gz, .fasta.gz) is supported.

A Dorado model string. The workflow supports base models and modification models:

Supported modification codes: 5mCG_5hmCG, 5mC, 6mA, 5mC_5hmC, 4mC_5mC

Run nextflow run shians/dorado_workflow --help to list all valid model strings (requires Nextflow ≥ 25.10.0).

*Exactly one of --dna or --cdna must be specified.

Resource limits are defined by process labels. To override them, create a custom.config file and pass it with -c custom.config.

Example custom.config:

process {
    withLabel: large {
        cpus   = 24
        memory = '128.GB'
        time   = '48h'
    }
    withLabel: medium {
        cpus   = 16
        memory = '64.GB'
        time   = '24h'
    }
}

Pass the custom config alongside a profile:

nextflow run shians/dorado_workflow \
  -profile gpu_wehi \
  -c custom.config \
  --pod5_sheet pod5_sheet.tsv \
  --reference_genome /path/to/genome.fa \
  --basecall_model "[email protected]" \
  --dna

nextflow run shians/dorado_workflow \
  -profile gpu_wehi \
  --pod5_sheet pod5_sheet.tsv \
  --reference_genome /path/to/genome.fa \
  --basecall_model "[email protected]" \
  --dna

nextflow run shians/dorado_workflow \
  -profile gpu_wehi \
  --pod5_sheet pod5_sheet.tsv \
  --reference_genome /path/to/genome.fa \
  --basecall_model "[email protected]" \
  --cdna

nextflow run shians/dorado_workflow \
  -profile gpu_wehi \
  --pod5_sheet pod5_sheet.tsv \
  --reference_genome /path/to/genome.fa \
  --basecall_model "[email protected]_5mCG_5hmCG@v2" \
  --dna

nextflow run shians/dorado_workflow \
  -profile gpu_wehi \
  --pod5_sheet pod5_sheet.tsv \
  --reference_genome /path/to/genome.fa \
  --basecall_model "[email protected]" \
  --dna \
  -resume

nextflow run shians/dorado_workflow \
  -profile gpu_wehi \
  --pod5_sheet pod5_sheet.tsv \
  --reference_genome /path/to/genome.fa \
  --basecall_model "[email protected]" \
  --dna \
  --output_dir /scratch/project/basecalled

nextflow run shians/dorado_workflow \
  -profile gpu_wehi \
  --pod5_sheet pod5_sheet.tsv \
  --reference_genome /path/to/genome.fa \
  --basecall_model "[email protected]" \
  --dna \
  --publish_fastq false

output/                        # Controlled by --output_dir
├── basecalled/                # Merged BAM files (one per sample)
│   ├── sample1.bam
│   └── sample2.bam
├── fastq/                     # Compressed FASTQ files (one per sample)
│   ├── sample1.fastq.gz
│   └── sample2.fastq.gz
└── logs/
    └── runtime_reports/
        ├── main_timeline.html # Interactive execution timeline
        ├── main_report.html   # Execution summary report
        └── main_trace.txt     # Detailed task trace log

BAM files include alignment tags from minimap2, making them suitable for downstream tools that use alignment information (e.g., splice site analysis, modification calling).

• Dorado
• Nextflow Documentation