Deep learning-based discovery of an antimicrobial peptide from Tenebrio molitor active against Gram-positive and Gram-negative bacteria

Yellow mealworm ( Tenebrio molitor ) is an EU-approved novel food and a promising source of edible protein, yet its potential to release antimicrobial peptides (AMPs) upon gastrointestinal digestion remains largely unexplored. In this study, we employed a deep learning (DL)-driven approach to systematically mine the T. molitor proteome for novel AMPs generated during gastrointestinal digestion. A comprehensive in silico peptide library of 3,782,689 fragments was constructed via virtual enzymatic digestion of 49,132 proteins. A tailored Transformer-based model was applied for rapid screening, achieving 92% accuracy and high specificity, which led to the identification of seven candidate AMPs. These peptides were synthesized, and their antimicrobial activity was validated in vitro against five foodborne pathogens ( Staphylococcus aureus, Salmonella typhimurium, Escherichia coli, Pseudomonas aeruginosa , and Listeria monocytogenes ). Among them, AP2 exhibited potent activity, with MIC values as low as 1/16th that of the commercial preservative Nisin Z. Further analysis revealed AP2’s ability to self-assemble into stable nanoparticles and its primary mechanism of action: disrupting bacterial membrane integrity, as confirmed by SEM and all-atom molecular dynamics simulations. Furthermore, AP2 demonstrated favorable thermal, pH, and salt stability, as well as low cytotoxicity against mammalian cells. This study utilized a deep learning framework to identify antimicrobial peptides from yellow mealworm, particularly the AP2 variant with potent antibacterial activity. It holds potential for future applications in food preservation and infection control.