AI-Driven Evolution of Antimicrobial Peptides Against Superbugs

A new AI-powered deep learning model, EvoGradient, has successfully identified and optimized antimicrobial peptides (AMPs) to combat multidrug-resistant pathogens. This explainable AI system not only predicts the potency of AMPs but also virtually evolves them to enhance their effectiveness, mimicking directed evolution in silico. By applying EvoGradient to peptides derived from low-abundance human oral bacteria, researchers generated 32 optimized AMPs, six of which were synthesized and tested. These peptides exhibited strong activity against deadly pathogens, including carbapenem-resistant E. coli, K. pneumoniae, and A. baumannii, as well as vancomycin-resistant E. faecium. Among them, pep-19-mod emerged as the most potent, demonstrating over 95% bacterial reduction in infected mouse models. This breakthrough highlights AI’s transformative potential in accelerating antibiotic discovery and development, offering a promising solution to the global antibiotic resistance crisis.