A generative artificial intelligence approach for peptide antibiotic optimization

2026-05-13 · Source: Nature Machine Intelligence · Field: Science & Research — Artificial Intelligence & Machine Learning, Life Sciences & Biology, Health & Medical Research · Depth: Expert, extended

Summary

ApexGO, a generative artificial intelligence framework, has been developed to optimize peptide antibiotics, addressing the global rise of antibiotic resistance. This system integrates a transformer variational autoencoder (VAE) to embed peptide sequences in a continuous latent space with Bayesian optimization (BO) to efficiently propose sequence edits. Using ten de-extinct peptides as templates, ApexGO generated optimized derivatives, with 100 compounds chemically synthesized and characterized in vitro. The system achieved an 85% experimental hit rate and a 72% success rate in enhancing antimicrobial activity against Gram-negative pathogens, outperforming previous methods. In preclinical mouse models of *Acinetobacter baumannii* infection, AI-optimized molecules demonstrated potent anti-infective activity, superior to template controls and comparable to or exceeding last-resort antibiotics like polymyxin B.

Key takeaway

For AI Scientists and Research Scientists focused on drug discovery, ApexGO offers a robust framework to accelerate antibiotic development. Its ability to optimize peptide scaffolds under specific constraints and achieve high experimental hit rates means you can efficiently generate potent antimicrobial candidates. Consider applying this generative AI and Bayesian optimization approach to your lead optimization efforts, especially for challenging targets like multidrug-resistant Gram-negative pathogens, to potentially reduce development time and improve efficacy.

Key insights

ApexGO uses generative AI and Bayesian optimization to efficiently design and optimize peptide antibiotics from existing templates.

Principles

• Transform discrete optimization problems into continuous ones via VAEs.
• Incorporate feedback loops for iterative design improvement.
• Utilize trust regions to manage exploration in high-dimensional latent spaces.

Method

ApexGO employs a transformer-based VAE to map peptide sequences into a continuous latent space, then uses Bayesian optimization, including local latent Bayesian optimization (LOL-BO) and multiple trust regions, to propose sequence edits for enhanced antimicrobial potency.

In practice

• Optimize existing peptide scaffolds under practical design constraints.
• Generate derivatives with specific similarity thresholds (e.g., ">=75%").
• Target specific pathogen strains for tailored antimicrobial activity.

Topics

• Generative AI
• Peptide Antibiotics
• Antimicrobial Resistance
• Bayesian Optimization
• Transformer VAE

Code references

• Yimeng-Zeng/ApexGO

Best for: AI Scientist, Research Scientist, Machine Learning Engineer

Related on AIssential

• Data-driven surrogates of rational design enable antimicrobial peptide optimization — Generative AI and data-driven surrogates enable rapid optimization of antimicrobial peptides against drug resistance.
• Evolutionary Generative Optimization: Towards Fully Data-Driven Evolutionary Optimization via Generative Learning — EvoGO leverages generative learning to create a fully data-driven evolutionary optimization framework, eliminating handcrafted heuristics.
• Pepti-Agent: An AI Agent for Peptide Design and Optimization — Pepti-Agent uses an LLM controller and inspectable tools for multi-objective peptide design guided by live property predictions.
• OMEGA: Optimizing Machine Learning by Evaluating Generated Algorithms — OMEGA automates ML algorithm generation, outperforming baselines through meta-prompt engineering and code generation.
• 🔬Beyond AlphaFold: How Boltz is Open-Sourcing the Future of Drug Discovery — AlphaFold's legacy drives open-source tools for democratizing advanced molecular structure prediction and generative design.
• The scientist using AI to hunt for antibiotics just about everywhere — AI-driven peptide discovery from diverse biological sources offers a novel approach to combat antimicrobial resistance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Nature Machine Intelligence.