AI agent-based discovery of antimicrobial peptides against multidrug-resistant gram-negative bacterial infection.

Abstract

The emergence of multidrug-resistant gram-negative bacteria poses a severe threat to global public health, and the limitations of traditional antibiotics in efficacy and drug resistance have become increasingly prominent. This study integrated AI technologies, including RFdiffusion and ProteinMPNN, to design and screen a novel antimicrobial peptide, AMP-ZJLC586, which exhibited potent antimicrobial activity against multidrug-resistant gram-negative bacteria, with a minimum inhibitory concentration (MIC) ranging from 0.5 to 8 μg/mL, and has the characteristics of high stability and low cytotoxicity. In addition, it can also kill clinically isolated drug-resistant bacteria and inhibit the formation of their biofilms. In terms of bactericidal mechanism, AMP-ZJLC586 can reduce the production of bacterial ATP by inhibiting the synthesis of lipopolysaccharide and the enzyme activity of respiratory chain, and oxidative stress response and SOS response at the same time. In the sepsis mouse model, AMP-ZJLC586 significantly increased the survival rate by 65%, effectively suppressed the level of inflammatory factors, improved organ damage, and reduced the bacterial load of target organs by 103-104 CFU/g, highlighting its great therapeutic potential.