TLDR: Researchers at the University of Pennsylvania have developed a generative AI model, AMP-Diffusion, capable of designing novel antibiotics from scratch. This AI tool generated tens of thousands of antimicrobial peptides (AMPs), with two leading candidates demonstrating efficacy comparable to FDA-approved drugs in animal models, without observed side effects. This breakthrough marks a significant step in combating the global crisis of antibiotic resistance.
In a groundbreaking advancement in the fight against antibiotic resistance, researchers at the University of Pennsylvania have unveiled a generative artificial intelligence (AI) model that can design potent new antibiotics. Published in the journal Cell Biomaterials, this innovative tool, named AMP-Diffusion, represents a paradigm shift from identifying existing antibiotic candidates to inventing entirely new molecules.
The collaborative effort, led by the labs of Assistant Professor Pranam Chatterjee and Presidential Associate Professor César de la Fuente, both from Penn Engineering, utilized a ‘diffusion model’ – similar to those used in generative AI for images like DALL-E and Stable Diffusion. Instead of refining pixels, AMP-Diffusion refines sequences of amino acids to create antimicrobial peptides (AMPs), which are short protein building blocks with bacteria-killing potential.
“We’re leveraging the same AI algorithms that generate images, but augmenting them to design potent new molecules,” stated Pranam Chatterjee, who is also an assistant professor in computer and information science. The model successfully generated approximately 50,000 peptide sequences. To manage this vast number, the researchers employed another AI tool, APEX 1.1, developed in de la Fuente’s lab, which ranked candidates based on predicted bacteria-killing power, novelty, and diversity.
From this extensive pool, 46 of the most promising AMPs were synthesized and rigorously tested in human cells and animal models. The results were remarkable: two of these AI-designed molecules demonstrated efficacy on par with FDA-approved antibiotics, specifically levofloxacin and polymyxin B, in treating skin infections in mice. Crucially, these potent new compounds showed no detectable adverse effects. “It’s exciting to see that our AI-generated molecules actually worked,” Chatterjee added, emphasizing that “This shows that generative AI can help combat antibiotic resistance.”
Also Read:
- Generative AI Poised to Revolutionize Clinical Study Report Development, Addressing Stagnant Timelines
- Cybercriminals Leverage Generative AI, Including ChatGPT, to Circumvent Antivirus Defenses
César de la Fuente highlighted the broader implications, noting, “Nature’s dataset is finite; with AI, we can design antibiotics evolution never tried.” This research not only demonstrates the capability of AI to invent new drugs but also promises to drastically accelerate the antibiotic discovery timeline, potentially reducing it from years to mere days. This efficiency is vital in addressing the escalating global threat of drug-resistant superbugs.
