Particle.news

Download on the App Store

Penn AI Designs Novel Peptide Antibiotics That Match Approved Drugs in Mice

Peer-reviewed results show a diffusion model grounded in Meta’s ESM-2 can yield mouse-active antibiotic peptides.

Overview

  • University of Pennsylvania researchers report the AMP-Diffusion system in Cell Biomaterials, published September 2, 2025.
  • The model generated roughly 50,000 antimicrobial peptide candidates, then an in‑lab tool called APEX 1.1 prioritized sequences for synthesis.
  • Forty-six peptides were made and tested, and two cleared skin infections in mice with efficacy comparable to levofloxacin and polymyxin B.
  • The peptides showed no detectable adverse effects in the animal studies, according to the authors.
  • The team frames the work as proof of principle and plans to steer designs toward drug-like properties, with extensive preclinical development still required.