Duet

Duet performs structural variant (SV) calling, genotyping, and phasing from Oxford Nanopore Technologies (ONT) MinION long-read sequencing data by integrating structural variant (SV) and single-nucleotide polymorphism (SNP) signatures to improve accuracy across sequencing coverages.

Key Features:

• Integration of SV and SNP signatures: Combines structural variant signatures with SNP signatures to distinguish true SV haplotypes from false signals.
• SV calling, genotyping, and phasing: Provides calling, genotype assignment, and haplotype phasing of structural variants from ONT long reads.
• Benchmarked performance across coverages: Evaluated against state-of-the-art tools on multiple ONT sequencing datasets across coverage levels from 8× to 40×.
• Superior low-coverage accuracy: Demonstrates improved performance in SV calling, genotyping, and phasing at low coverage (8×) compared to other tools.
• Improved phasing at higher coverage: Maintains superior SV calling and genotyping and shows increased F1-score for SV phasing as coverage increases to 20× and up to 40×.
• Applicability to clinical scenarios: Enables accurate analyses with low-coverage data for cost-sensitive clinical applications while scaling to high-coverage research genomes.

Scientific Applications:

• Structural variant discovery and characterization: Precise identification and characterization of SVs from ONT long-read data.
• Clinical diagnostics: Support for diagnostic workflows that require SV detection and phasing, including low-coverage clinical sequencing.
• Genetic diversity and population studies: Analysis of structural variation contributing to population-level genetic diversity.
• Disease mechanism investigation: Linking phased SVs to haplotypes to study mechanisms of genetic disease.
• Personalized medicine: Improving accuracy of genomic data interpretation for individualized diagnostic and therapeutic decisions.

Methodology:

Duet uses algorithms that integrate ONT long-read sequencing data with SNP information in a dual-signature approach to discriminate true structural variants from noise and to support SV calling, genotyping, and phasing.