tools-only Basecalling - Usage Guide
Overview
Basecalling converts raw Nanopore signal data (FAST5/POD5) to nucleotide sequences. Dorado is the current production basecaller and should be used for all new analyses. Guppy is deprecated and no longer receiving updates.
Prerequisites
# Dorado (from ONT community)
# Download from https://github.com/nanoporetech/dorado
# POD5 tools
pip install pod5
# Quality filtering
conda install -c bioconda chopper nanoplot
Quick Start
Tell your AI agent what you want to do:
- "Basecall my POD5 files with Dorado using SUP accuracy"
- "Basecall my FAST5 files with Dorado"
Example Prompts
Basic Basecalling
"Basecall my POD5 files in the raw_data/ folder using Dorado with the SUP model"
"Run Dorado on my R10.4.1 data with HAC accuracy for faster processing"
Model Selection
"What Dorado model should I use for R10.4.1 chemistry?"
"Basecall with the fast model for a quick preview of my data"
With Modifications
"Basecall and detect 5mC methylation using Dorado"
Quality Filtering After
"Basecall my data and then filter for Q10+ reads"
What the Agent Will Do
- Identify your flow cell chemistry (R10.4.1 vs R9.4.1)
- Select appropriate model based on speed/accuracy tradeoff
- Run basecalling with Dorado
- Optionally filter output for quality
Model Selection
Choose based on speed vs accuracy tradeoff:
- fast: Quick preview, lowest accuracy
- hac: Balanced, good for most uses
- sup: Highest accuracy, slowest
Chemistry Matching
Match the model to your flow cell chemistry:
- R10.4.1: Current chemistry
- R9.4.1: Legacy chemistry
Workflow Position
Raw Signal (FAST5/POD5)
|
v
Basecalling (Dorado)
|
v
FASTQ/BAM
|
v
Quality Filtering (chopper)
|
v
Alignment (minimap2)
Tips
- Use SUP model for final analysis, HAC for exploratory work, fast for quick checks
- Dorado can output BAM directly with alignment if reference provided
- Use
--batchsize 32if running out of GPU memory - Check chemistry (R10.4.1 vs R9.4.1) before selecting model
- POD5 is the new format; FAST5 is legacy but still supported