A team of researchers from the United States applied artificial intelligence (AI) to develop complete genomes of new bacteriophages. This approach marks the first attempt of its kind and is a step toward potentially creating AI-designed phage therapies, according to research published in the journal Science on Thursday.
The researchers, led by principal investigator and corresponding author Brian Hie, an assistant professor at Stanford University and innovation researcher at Arc Institute, utilized two genome language models—Evo 1 and Evo 2. These models allowed for the generation of complete phage genomes possessing realistic genetic architecture and specificity to the bacterial host Escherichia coli C. (E. coli). The researchers noted that this was the first generative development of complete bacteriophage genomes.
Bacteriophages, also known as phages, are viruses that infect bacteria, not humans, animals, or plants. According to the study, scientists generated and filtered thousands of sequence variants before selecting about 300 designs.
In laboratory tests, 285 of these designs were successfully synthesized and assembled in E. coli cells, and 16 of them demonstrated the ability to form viable bacteriophages that replicated and destroyed bacteria.
Some of the designed phages showed higher efficacy in destroying bacteria in a direct laboratory competition compared to the natural phage phiX174, which infects E. coli. Furthermore, a mixture of the developed phages proved effective against E. coli strains resistant to phiX174, indicating potential for AI-created phage therapies.
In a companion 'Perspective' article, Thomas Inglesby and Moritz Hanke from Johns Hopkins University expressed concern regarding the biosecurity implications of this achievement. They called for a legal requirement to screen orders for synthetic genetic material for potentially dangerous sequences instead of relying solely on voluntary safeguards.
