Scientists at Stanford University have, for the first time, used artificial intelligence to design viruses not found in nature that could kill bacteria resistant to natural bacteriophages, several international media outlets reported on Friday.
Natural bacteriophages are already used to treat infections when antibiotics no longer work. However, bacteria can also become resistant to those viruses.
In laboratory tests, a mix of AI-designed bacteriophages successfully killed E. coli strains that had resisted natural phages.
To create them, the researchers used AI to design genomes, the complete set of hereditary material in an organism, for new bacteriophages. They relied on genomic language models, a genetic version of the language models that underpin conversational AI systems.
Writing in the Science journal, the researchers said their approach broadens the scope of synthetic genomics and could open the way to phage therapies that can adapt to rapidly evolving pathogens.
Biosafety concerns
The study also comes amid growing concern that AI could one day be used to create dangerous pathogens affecting humans or animals.
In a commentary published in the same issue of Science, Professor Tom Inglesby and researcher Moritz Hanke of the Johns Hopkins Center for Health Security said the technology shows promise for the life sciences but raises urgent biosafety questions.
They wrote that assembling viral genomes using generative AI is now possible, yet policies to safely regulate the practice are still lacking.
Professor Tom Ellis of Imperial College London said, however, that bacteriophages are among the easiest genomes to reproduce. He added that artificially creating more complex genomes is not yet on the horizon.
"The threat of a virus or bacterium genome being fully designed by AI is greatly exaggerated," he said.
