ExtrARG

ExtrARG identifies discriminatory antibiotic resistance genes (ARGs) within environmental resistomes using an extremely randomized tree (ERT) algorithm combined with Bayesian optimization to detect environment-specific ARG occurrence patterns.

Key Features:

• Extremely Randomized Tree with Bayesian Optimization: Applies an extremely randomized tree (ERT) algorithm integrated with Bayesian optimization to capture ARG variability and select discriminatory features.
• Non-biased Feature Selection: Identifies discriminatory ARGs based on their discriminatory signal rather than high relative abundance, reducing abundance-driven bias.
• Validation on Simulated Data: Validated using simulated metagenomic Illumina sequencing data to assess discriminatory ARG identification performance.
• Application to Diverse Aquatic Datasets: Applied to publicly available and proprietary metagenomic datasets from aquatic habitats including river, wastewater influent, hospital effluent, dairy farm effluent, and river samples from the Amazon, Kalamas, and Cam Rivers.
• Comparative Resistome Analysis: Enables comparison and ranking of resistomes between distinct or similar environments to characterize variance in ARG profiles.
• Surveillance and Mitigation Assessment: Identifies discriminatory ARGs according to predefined categorizing schemes to support ARG surveillance and assess mitigation strategies.

Scientific Applications:

• Environmental Monitoring: Detects environment-specific ARGs in environmental compartments to monitor ARG proliferation across habitats.
• Comparative Analysis: Supports comparative analyses and ranking of resistomes across environments to evaluate potential contributions to clinically relevant antibiotic resistance.
• Informing Interventions: Identifies key discriminatory ARGs to inform targeted interventions and mitigation policies aimed at controlling antibiotic resistance spread.

Methodology:

Integrates an extremely randomized tree (ERT) algorithm with Bayesian optimization and was validated on simulated metagenomic Illumina sequencing data and applied to metagenomic datasets from aquatic habitats (river, wastewater influent, hospital effluent, dairy farm effluent, and river samples from the Amazon, Kalamas, and Cam Rivers).