Protein Language Model‐Driven Optimisation of Antimicrobial Peptide Pth‐Ca1 Against <i>Pectobacterium brasiliense</i> Using <scp>ESMFold</scp> ‐Predicted Structures and the <scp>ESM</scp> ‐3 Model

The emergence of antimicrobial resistance (AMR) poses a significant threat to global health and food security. Antimicrobial peptides (AMPs), particularly those characterised by α-helical structures, represent a promising alternative due to their broad-spectrum activity and unique mechanisms of action. Pectobacterium brasiliense is a destructive bacterial pathogen affecting solanaceous crops but effective control measures remain insufficient. This study aims to optimise pseudothionin AMPs using AI-based protein language models, ESM-3 and ESMFold, to enhance their antibacterial efficacy. Using ESM-3, we generated multiple analogs of Pth-Ca1 with increased net charge, hydrophobicity and helical ratio. Among these, Design_1867 exhibited the strongest antibacterial activity against both Escherichia coli and P. brasiliense, with minimum inhibitory concentration (MIC) values of 31.25 μg/mL. Design_1867 was found to bind bacterial DNA and induce pore formation in bacterial membranes through a barrel-stave mechanism, similar to the reference peptide alamethicin. Reverse transcription-quantitative PCR analyses revealed the downregulation of key genes associated with membrane integrity and biofilm formation. In planta assays confirmed its efficacy and low cytotoxicity. This study demonstrates the successful application of ESM-3 and ESMFold for the rational design of highly effective AMPs. Design_1867 exhibits potent antimicrobial activity against P. brasiliense with minimal toxicity, underscoring the potential of AI-driven AMP optimisation for sustainable agricultural disease management.