XBS

XBS performs joint variant calling and machine-learning-based filtering to detect single nucleotide polymorphisms (SNPs) and insertions/deletions (indels) in Mycobacterium tuberculosis whole-genome sequencing data from contaminated and low-coverage clinical samples.

Key Features:

• Joint variant calling and machine-learning-based filtering: Integrates joint variant calling with machine-learning filtering to enable accurate variant detection from unbiased sputum samples and other challenging inputs.
• Handling of complex genomic regions: Accurately detects SNPs and indels in complex, repetitive regions and reports 9.0% more SNPs and 8.1% more indels compared to the WHO-endorsed unified analysis variant pipeline.
• Robustness to low coverage and high contamination: Performs variant detection on sputum-derived sequence data with low depth (5–10×) and remains effective regardless of contamination type or levels exceeding 50%.
• High sensitivity and specificity: Achieves 98.8% sensitivity on culture isolates, identifies 13.9% more variable sites in sputum samples than MTBseq, and produces no false positives when ribosomal RNA regions are excluded.
• Generalisability to other bacteria: Designed to be applicable to analysis of other complex bacterial genomes beyond Mycobacterium tuberculosis.

Scientific Applications:

• Drug resistance and virulence studies: Enables sequencing of contaminated and low-coverage clinical M. tuberculosis specimens to characterize genomic variants associated with drug resistance and virulence.
• Genome-wide association studies (GWAS): Provides increased genetic resolution that enhances the discovery potential in GWAS of M. tuberculosis.
• Transmission studies: Improves variant detection for sequence-based transmission and epidemiological investigations of tuberculosis.

Methodology:

XBS performs joint variant calling combined with machine-learning-based filtering and excludes ribosomal RNA regions during variant inference to produce high-quality SNP and indel calls.