PARGT

PARGT predicts antimicrobial-resistance (AMR) genes in bacterial genomes using machine-learning classification informed by a game-theory-based feature evaluation.

Key Features:

• Machine-learning integration: Employs machine-learning techniques to identify AMR genes without relying on traditional alignment-based methods.
• Game-theory-based feature evaluation: Uses a game-theory-based algorithm to identify critical protein features that inform the machine-learning model.
• Broad applicability and performance: Validated on Gram-negative bacteria and extended to Gram-positive bacteria, with reported classification accuracies of 87%–90% for genes encoding resistance to bacitracin and vancomycin.

Scientific Applications:

• AMR gene discovery: Predicts novel and known antimicrobial-resistance genes from bacterial genomic data.
• Resistance mechanism research: Identifies protein features relevant to mechanisms of bacterial antibiotic resistance.
• Antibiotic development support: Provides candidate resistance determinants useful for guiding antibiotic design and testing.
• Public-health surveillance: Aids detection and classification of resistance genes across Gram-negative and Gram-positive bacteria for surveillance efforts.

Methodology:

Feature identification using a game-theory-based algorithm to select essential protein features, followed by machine-learning classification that uses those features to predict the presence of AMR genes.