PARGT

PARGT (Predictive Antimicrobial Resistance Gene Tool) is software that enhances the identification of new antimicrobial-resistance genes using machine-learning approaches. Leveraging the vast availability of whole-genome sequences, PARGT addresses the limitations of traditional alignment-based methods, especially when pathogens cannot be cultured in the laboratory. This tool is based on a game-theory-based feature evaluation algorithm that identifies specific protein characteristics (features) crucial for machine learning models to identify antimicrobial resistance (AMR) genes accurately.

The PARGT methodology, initially designed for Gram-negative bacteria, has been expanded to include Gram-positive bacteria, highlighting its versatility and effectiveness across different types of pathogens. The software utilizes advanced machine learning techniques with features identified through game theory to achieve notable classification accuracies in 87% to 90% for genes that encode resistance to antibiotics such as bacitracin and vancomycin.