Generative latent diffusion language modeling yields anti-infective synthetic peptides
Marcelo D T Torres1,2,3,4, Tianlai Chen5, Fangping Wan1,2,3,4
1Machine Biology Group, Departments of Psychiatry and Microbiology, Institute for Biomedical Informatics, Institute for Translational Medicine and Therapeutics, Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania, United States of America.
Biorxiv : the Preprint Server for Biology
|February 20, 2025
Summary
Generative artificial intelligence rapidly discovers novel antimicrobial peptides (AMPs) using the AMP-Diffusion model. These AI-designed peptides show broad-spectrum activity against resistant bacteria with low toxicity, offering a promising strategy against antimicrobial resistance.
Area of Science:
- Biotechnology
- Computational Biology
- Drug Discovery
Background:
- Antimicrobial peptide (AMP) design faces challenges due to vast sequence space and complex structure-activity relationships.
- Traditional methods struggle to efficiently navigate peptide diversity for antibiotic discovery.
- Antimicrobial resistance necessitates novel therapeutic strategies.
Purpose of the Study:
- To introduce AMP-Diffusion, a novel generative artificial intelligence (AI) model for accelerated antimicrobial peptide design.
- To discover and validate novel antibiotic candidates with broad-spectrum activity and low toxicity.
- To demonstrate the potential of AI-driven platforms in addressing antimicrobial resistance.
Main Methods:
- Developed AMP-Diffusion, a latent diffusion model fine-tuned on AMP sequences using protein language model embeddings.
- Generated 50,000 candidate sequences, filtered and ranked using the APEX predictor model.
- Synthesized and experimentally validated 46 top peptide candidates, including in vitro assays and preclinical mouse models.
Main Results:
- AMP-Diffusion successfully generated promising antibiotic candidates.
- Validated peptides exhibited broad-spectrum antibacterial activity against pathogens, including multidrug-resistant strains.
- Lead peptides demonstrated low cytotoxicity, effective bacterial killing via membrane disruption, and efficacy in mouse infection models comparable to existing antibiotics.
Conclusions:
- AMP-Diffusion is a robust generative platform for designing novel antimicrobial peptides and antibiotics.
- AI-driven peptide design offers a promising strategy to combat the growing threat of antimicrobial resistance.
- The developed peptides show potential as effective therapeutics with favorable safety profiles.
