AI agent-based discovery of D-enantiomeric antimicrobial peptides against multidrug-resistant bacterial infection.

Antimicrobial peptides (AMPs) offer a route to counter resistant pathogens but are often hampered by proteolysis, whereas D-peptides resist proteases yet remain underexplored due to data scarcity and design challenges. Here, we present PeptiD-Agent, a purely agent based framework that predicts D-peptide antimicrobial activity with extremely limited data, enabling rapid discovery of potent candidates. Using this approach, we identified DA2, a D-enantiometric AMP lead with broad-spectrum activity against drug-resistant bacteria and minimal hemolytic toxicity. DA2 showed high stability under physiological conditions, including resistance to enzymatic degradation and serum. Mechanistic studies indicate that DA2 exerts bactericidal effects by disrupting the integrity of the bacterial membrane in concert with multiple synergistic mechanisms. In murine models of skin wounds and intraperitoneal infection, DA2 conferred significant protection against drug-resistant pathogens and, when delivered via hydrogel, accelerated wound healing. These findings establish a computational route to potent, stable D-peptide antimicrobials and provide a general strategy for AMP design in data-scarce settings.