Clair3

Overview

Clair3 is a deep-learning-based variant caller designed specifically for long-read sequencing data from Oxford Nanopore Technologies (ONT) and PacBio platforms. It uses a pileup-based neural network followed by a full-alignment model to call germline SNPs and indels with high accuracy. Clair3 ships with pre-trained models for various sequencing platforms and chemistries, making it straightforward to deploy on both Nanopore simplex and duplex data as well as PacBio HiFi reads. It supports multi-threaded execution and can process whole-genome datasets efficiently.

Installation

mamba create -n clair3 -c bioconda clair3

Activate the environment before running:

conda activate clair3

Basic Usage

Call variants from ONT reads aligned to a reference genome.

run_clair3.sh \
  --bam_fn=aligned.sorted.bam \
  --ref_fn=reference.fna \
  --output=clair3_output/ \
  --threads=8 \
  --platform="ont" \
  --model_path="${CONDA_PREFIX}/bin/models/ont" \
  --sample_name=sample1 \
  --include_all_ctgs

Key Parameters

Flag / option	Description
`--bam_fn`	Path to the sorted and indexed BAM file.
`--ref_fn`	Path to the reference FASTA file (must be indexed).
`--output`	Output directory for variant calls and intermediate files.
`--threads`	Number of CPU threads to use.
`--platform`	Sequencing platform: `ont` for Oxford Nanopore, `hifi` for PacBio HiFi, or `ilmn` for Illumina.
`--model_path`	Path to the pre-trained model directory matching the sequencing platform and chemistry.
`--sample_name`	Sample name to embed in the VCF header.
`--include_all_ctgs`	Call variants on all contigs, not just those matching `chr` naming conventions.
`--bed_fn`	Restrict variant calling to regions defined in a BED file.
`--qual`	Minimum variant quality score threshold (default 2).

Expected Output

Clair3 writes its results to the specified output directory:

merge_output.vcf.gz – the final merged VCF containing all called SNPs and indels with quality scores and genotype information.
merge_output.vcf.gz.tbi – tabix index for the merged VCF.
pileup.vcf.gz – intermediate pileup model calls before full-alignment refinement.
full_alignment.vcf.gz – calls from the full-alignment model for candidate variants.
log/ – directory containing run logs and timing information.