AI Decodes Superbug Defences from DNA with 99% Accuracy

Antimicrobial resistance is a major global health threat, as pathogens continually evolve new ways to defeat our medicines. To combat this, researchers have developed a promising new tool that uses machine learning to predict a microbe's defensive capabilities directly from its genetic code.

The project, named ASAP, tested ten different machine learning models, training them on pathogen gene sequences. The goal was to see how accurately they could predict whether a pathogen would be resistant to a specific antibiotic. The results were striking: the top-performing model, XGBoost, achieved an accuracy of 0.99. Even the least predictive model still reached a respectable 0.89 accuracy.

This method has the potential to improve how 'antibiograms'—charts showing which drugs are effective—are created. By providing healthcare professionals with rapid, data-driven insights, this AI-powered approach could lead to more precise antibiotic prescribing and reduce the guesswork often involved in treatment.