Abstract
Copyright © 2025. Published by Elsevier Ltd.Antimicrobial peptides (AMPs) offer promising alternatives to conventional antibiotics due to their broad-spectrum activity and multi-modal mechanisms. However, large-scale experimental screening remains costly and time-consuming. To expedite AMP discovery, this study proposes a combined screening strategy based on Graph Attention Networks and molecular dynamics simulations. Utilizing a food-derived bioactive peptide database, the method efficiently identified candidate AMPs with activity against Escherichia coli ATCC 25922 and methicillin-resistant Staphylococcus aureus ATCC 33591 from a pool of 100,663,296 peptides. Experimental validation confirmed that the identified AMP, AMP_321, exhibited broad-spectrum antimicrobial activity. The peptide demonstrated excellent stability and low hemolytic activity. Mechanistic studies revealed that AMP_321 exerts its antibacterial effect by disrupting bacterial membrane integrity, inducing extracellular ATP release, and promoting intracellular accumulation of reactive oxygen species. In food models, AMP_321 significantly reduced bacterial contamination on beef and tomato surfaces, validating its potential as a safe and effective natural food preservative.
NETL
NSTL
Elsevier