First-of-Its-Kind Antibiotic-Resistant Bacteria Drugs Developed Using Generative AI
Researchers at Stanford Medicine and McMaster University are addressing the global challenge of antibiotic resistance using generative artificial intelligence.
The new model, SyntheMol, designed six potential drugs targeting resistant strains of Acinetobacter baumannii, a significant contributor to antibacterial resistance-related deaths.
Described in a study published in Nature Machine Intelligence, SyntheMol creates structures and chemical recipes for these compounds.
The potential of AI to design entirely novel molecules, previously undiscovered in nature.
Before the emergence of generative AI, researchers employed various computational methods for antibiotic development, such as scanning existing drug libraries.
Despite analyzing 100 million compounds, this approach only scratched the surface of the vast chemical space potentially housing antibacterial agents, as explained by Kyle Swanson, a co-lead author of the study.
Generative AI, though prone to generating impractical compounds, holds promise for drug discovery. To ensure feasibility, the researchers guided SyntheMol to create molecules synthesizable in a lab.
The model, trained on a library of over 130,000 molecular building blocks and validated chemical reactions, not only generated compounds but also provided synthesis instructions.
Additionally, it incorporated antibacterial activity data against A. baumannii to filter out compounds resembling existing ones, aiming to impede resistance development.
In less than nine hours, SyntheMol produced around 25,000 potential antibiotics and synthesis recipes, demonstrating its efficiency in generating diverse drug candidates.
