Artificial intelligence (AI) has been able to design tens of thousands of candidate antimicrobial peptides (AMPs), some of which can perform as well as standard antibiotics.
The findings, in Cell Biomaterials , demonstrate the capacity of generative AI—which forms the basis of ChatGPT—to design antibiotics as yet undiscovered naturally through evolution.
Using the generative AI platform AMP-diffusion, researchers were able to create amino-acid sequences for 50,000 peptides. These were whittled down to the 46 most promising candidates, three quarters of which were able to inhibit bacteria human cells and animal models.
Of note, two of these had similar efficacy to drugs approved by the U.S. regulators to treat antibiotic-resistant bacteria.
“We’re leveraging the same AI algorithms t