“If bacteria develop resistance to one phage, the treatment is over,” said Brian Hie, a chemical engineer at Stanford University and co-author of a new study on viruses designed by artificial intelligence. His point is that a single virus is a fragile weapon against infection. A mixture of different ones is harder to beat.
Hie and his Stanford colleagues used an AI model called Evo 2, which generates novel DNA sequences, to create a series of bacteriophages. These are viruses that destroy bacteria. In laboratory tests, a mixture of 16 highly effective designs rapidly killed E. coli that had already become resistant to naturally occurring phages, Euronews reported.
Why a cocktail matters
E. coli is a group of bacteria that can cause intestinal infections, and it is increasingly resistant to available antibiotics. Resistance is growing worldwide, largely because antibiotics are used excessively and inappropriately, which pushes bacteria to develop ways of surviving them.
“But if a mixture contains several genetically distinct phages,” Hie said, “it will be much more difficult for bacteria to develop resistance to the entire cocktail.”
The researchers say the approach could eventually be turned on other dangerous bacteria, including the one that causes tuberculosis and healthcare-associated infections such as MRSA (methicillin-resistant Staphylococcus aureus).
An open model and the safety debate
The team has made Evo 2 openly and freely available, and the authors acknowledge that doing so has set off discussion about safety. The main worry is that malicious actors could adapt the technology to create dangerous biological agents.
Simon Clark, an associate professor of cellular microbiology at the University of Reading in the UK, was not involved in the study. “As the authors themselves point out, this raises serious regulatory and safety questions,” he said. He found it “somewhat reassuring” that these scientists had added safeguards and taken extra precautions, since research of this kind is normally strictly controlled. “But there is no guarantee that every other scientist who decides to pursue similar work will be equally cautious,” he added.
Clark also put the risk in proportion. Natural pathogens pose a much greater threat at present than AI-designed ones, he said, because they are easier to obtain and produce than entirely new agents built from scratch.
The authors argue that AI-designed biology has one advantage over natural evolution: safety mechanisms can be built directly into the design process.
Sources
Sources
- World's first AI-designed viruses a step towards AI-generated life
- Safety fears as scientists make first viruses designed by AI | Science
