Researchers at Stanford University have successfully designed functioning viral genomes using artificial intelligence, marking a significant development in biotechnology. The viruses created are bacteriophages, which target bacteria rather than human cells. In laboratory testing, a combination of these AI-designed bacteriophages effectively eliminated strains of E coli that had developed resistance to naturally occurring bacteriophages, demonstrating potential therapeutic applications.
Dr Brian Hie and his team employed genome language models—AI systems analogous to those powering large language model chatbots—to generate the viral genetic sequences. The researchers trained their AI systems, called Evo1 and Evo2, on genetic data from 2 million bacteriophages while deliberately excluding genetic information from viruses that infect plants, humans, or animals. The process generated thousands of potential genomes; nearly 300 were synthesized in laboratories, though only 16 proved viable. Despite the low efficiency rate, the resulting bacteriophage cocktail successfully overcame resistance in multiple E coli strains.
The accomplishment opens possibilities for advancing phage therapy and expanding biotechnological applications, according to findings published in the journal Science. However, the researchers emphasized the need for careful governance. They called on other scientists designing complete genomes to engage safety and security professionals throughout their work, noting the emergence of significant biosafety, biocontainment, and biosecurity considerations.
Experts from Johns Hopkins University and other institutions have amplified these concerns, noting that while the technology offers medical promise, governance frameworks to safely manage AI-driven viral genome design remain underdeveloped. They warned that applying similar techniques to viruses capable of infecting humans, animals, or plants could create pathogens resistant to existing containment measures. Specialists suggested implementing layered oversight approaches, including safeguards around AI model development, responsible research review processes, DNA synthesis screening, and established laboratory biosecurity protocols.
Article Attribution | Read More at Article Source
Article summary produced by Claude AI