Skip to content

Clair3 Specifications

Process Details

  • Name: Clair3
  • Process UUID: f931b9m4xn2p0ev6x74vgt9vhvljf9
  • Process Group: Laava

Overview

Clair3 is a germline small variant caller specifically designed for long-read sequencing data. This process combines two complementary variant calling approaches: pileup calling for efficient handling of most variant candidates, and full-alignment analysis for complex variants to maximize both precision and recall. The tool is optimized for Oxford Nanopore Technologies (ONT) and Pacific Biosciences (PacBio) long-read sequencing platforms.

This process is implemented in Bash, which invokes the Clair3 variant calling pipeline using pre-trained deep learning models.

Key Functionality

  • Germline Variant Detection: Identifies single nucleotide variants (SNVs) and small insertions/deletions (indels) from long-read sequencing data
  • Hybrid Calling Strategy: Combines pileup-based calling for speed with full-alignment analysis for accuracy on complex variants
  • Platform-Specific Models: Utilizes pre-trained deep learning models optimized for different sequencing platforms and basecalling models
  • Comprehensive Output: Generates VCF files with variant calls and quality scores for downstream analysis

Input/Output Specification

Inputs

Required Inputs

  • BAM/BAI Files

    • Description: Aligned long-read sequencing data with corresponding index files
    • Format: BAM with BAI index

  • Reference Genome

    • Description: Reference genome sequence in FASTA format for variant calling
    • Format: FASTA

Optional Inputs

  • Target Regions

    • Description: BED file specifying genomic regions to restrict variant calling
    • Format: BED

  • Known Variants

    • Description: VCF file containing known variants for comparison or filtering
    • Format: VCF

Outputs

  • Variant Calls
    • Description: Called variants in standard VCF format with quality scores and genotype information
    • Format: VCF

Parameters & Settings

These parameters can be adjusted in the Foundry UI when running this process.

  • Sequencing Platform

    • Description: Select the sequencing platform of the input. Possible options: {ont,hifi,ilmn}.
    • Available options: hifi (default), ont, ilmn

  • Pre-trained Model

    • Description: Select a pre-trained model to use in variant calling.
    • Available options: hifi_revio (default), hifi_sequel2, r941_prom_sup_g5014, r941_prom_hac_g360+g422, r941_prom_hac_g238, ilmn

  • Optional Parameters

    • Description: Optional parameters used in Clair3. Use =value for optional parameters, e.g., --sample_name=STR instead of --sample_name STR.
    • Default value: (empty)

References & Resources

  • Tool Documentation: Contact the team for details on run_clair3.sh
  • Related Papers: Zheng, Z., Li, S., Su, J., Leung, A.W., Lam, T.W., & Luo, R. (2022). Symphonizing pileup and full-alignment for deep learning-based long-read variant calling. Nature Computational Science, 2(12), 797-803.