aceA generative artificial intelligence approach for peptide antibiotic optimization
Abstract Antibiotic resistance is rising globally, demanding faster, more reliable routes to design antimicrobial candidates. Although artificial-intelligence-based methods have accelerated antimicrobial discovery, most are designed to screen fixed libraries or generate candidates broadly, rather than optimize existing peptide scaffolds under practical design constraints. Here, to address this challenge, we present APEX generative optimization (ApexGO). ApexGO uses a transformer variational autoencoder that embe…
"NMCSD Microbiology Laboratory" by NavyMedicine is marked with Public Domain Mark 1.0. To view the terms, visit https://creativecommons.org/publicdomain/mark/1.0/.
Researchers developed ApexGO, a generative AI approach that modifies existing peptide scaffolds using a transformer variational autoencoder and Bayesian optimization. Using ten template peptides, the system proposed optimized derivatives, 100 of which were synthesized and tested in vitro and in two mouse models of Acinetobacter baumannii infection.
The work matters because it moves AI antibiotic discovery from broad library screening to constrained optimization of known scaffolds, reporting high experimental validation rates and in vivo activity comparable to last-resort antibiotics. What remains uncertain is generalizability beyond the ten templates and Gram-negative pathogens tested, and whether efficacy and safety will translate beyond preclinical models.
- ApexGO uses a transformer variational autoencoder that embeds peptide sequences in a continuous latent space, whereas Bayesian optimization efficiently proposes sequence edits
- Using ten peptides as templates, researchers chemically synthesized 100 optimized compounds for in vitro characterization of activity, mechanism, structure and cytotoxicity
- Testing included two preclinical mouse models of Acinetobacter baumannii infection comparing AI-optimized molecules to template controls and last-resort antibiotic
ApexGO generated optimized peptide derivatives that showed enhanced antimicrobial activity in vitro and potent anti-infective efficacy in mouse models of Acinetobacter baumannii infection.
The rundown
The study describes ApexGO, which embeds peptide sequences in a continuous latent space and proposes edits via Bayesian optimization, generating new sequences through modifications of template peptides rather than screening fixed libraries.
From ten template peptides, 100 derivatives were synthesized and characterized for antimicrobial activity, mechanism of action, secondary structure and cytotoxicity, with reported 85% hit rate and 72% success in enhancing activity against Gram-negative pathogens, and superior efficacy to templates in two mouse models of Acinetobacter baumannii.
Sources
- Peer-reviewedNature Machine Intelligence2026-05-13
How should this claim be treated?
The debate