In a hopeful sign of the demand for safer, more effective antibiotics for humans, researchers at the University of Texas at Austin have used artificial intelligence to develop a new drug that is already showing promise in animal testing.
Publish their results in Nature Biomedical Technologythe scientists describe using a large language model – an AI tool like the one that powers ChatGPT – to develop a version of a bactericidal drug that was previously toxic to humans so that it would be safe to use.
The prognosis for patients with dangerous bacterial infections has worsened in recent years as antibiotic-resistant strains of bacteria spread and the development of new treatment options has stalled. However, UT researchers say that AI tools are groundbreaking.
“We have found that large language models are a major step forward for machine learning applications in protein and peptide engineering,” says Claus Wilke, professor of integrative biology and statistics and data sciences, and co-senior author of the new paper. “Many use cases that were not feasible with previous approaches are now starting to work. I foresee these and similar approaches being widely used to develop therapies or drugs in the future.”
Large language models, or LLMs, were originally designed to generate and explore sequences of text, but scientists are finding creative ways to apply these models to other domains. For example, just as sentences are made up of strings of words, proteins are made up of strings of amino acids. LLMs group words that share common features (such as cat, dog and hamster) into what is known as an ’embedding space’ with thousands of dimensions. Likewise, proteins with similar functions, such as the ability to fight dangerous bacteria without hurting the people harboring the bacteria, can cluster together in their own version of an AI embedding space.
“The space containing all the molecules is enormous,” says Davies, co-senior author of the new paper. “Machine learning allows us to find the regions of chemical space that have the properties we are interested in, and it can do this much faster and more thoroughly than standard laboratory approaches.”
For this project, the researchers used AI to find ways to re-engineer an existing antibiotic called Protegrin-1, which is good at killing bacteria but toxic to humans. Protegrin-1, which is naturally produced by pigs to fight infections, is part of a sub-type of antibiotics called antimicrobial peptides (AMPs). AMPs generally kill bacteria directly by disrupting cell membranes, but many of them target both bacterial and human cell membranes.
First, the researchers used a high-throughput method they had previously developed to create more than 7,000 variations of Protegrin-1 and quickly identify regions of the AMP that could be modified without losing antibiotic activity.
They then trained a protein LLM on these results so that the model could evaluate millions of possible variations for three traits: selectively targeting bacterial membranes, powerfully killing bacteria, and not damaging human red blood cells to find those that target the fell into a good place. three. The model then helped the team create a safer, more effective version of Protegrin-1, which they called bacterially selective Protegrin-1.2 (bsPG-1.2).
Mice infected with multidrug-resistant bacteria and treated with bsPG-1.2 were much less likely to have detectable bacteria in their organs six hours after infection, compared to untreated mice. If further testing yields similarly positive results, the researchers hope to eventually use a version of the AI-informed antibiotic in human trials.
“The impact of machine learning is twofold,” Davies said. “It’s going to identify new molecules that have the potential to help people, and it will show us how we can use and improve those existing antibiotic molecules and focus our work to get them into clinical practice more quickly.”
This project highlights how academic researchers are advancing artificial intelligence to meet societal needs, a major theme this year at UT Austin, which has declared 2024 the Year of AI.
The study’s other authors are research associate Justin Randall and graduate student Luiz Vieira, both at UT Austin.
Funding for this research was provided by the National Institutes of Health, The Welch Foundation, the Defense Threat Reduction Agency and Tito’s Handmade Vodka.
