Stanford Researchers Use Generative AI to Design Functional Bacteriophage Viruses

Researchers at Stanford University have used generative artificial intelligence to design complete and functional genomes for bacteriophages, which are viruses that infect bacteria. According to a study published in the journal Science, the team utilized genomic language models trained on public genomes from the Microviridae family. These models treat DNA nucleotides as words and phrases to design genetic material from scratch. The resulting synthetic DNA sequences were produced by commercial companies and tested in laboratories, where they were confirmed to be biologically viable and capable of multiplying within Escherichia coli C bacteria.

The researchers implemented several biosafety measures to prevent the creation of harmful pathogens. The AI was trained exclusively on bacteriophages, which cannot infect humans or other eukaryotic organisms, and data from human viruses were preemptively excluded from the training set. Additionally, the team used computational design rules known as tropism filters to ensure the viruses only attacked the specific host bacteria chosen for the test.

While the research offers potential for personalized therapies to combat antibiotic resistant bacteria, it also raises questions about security. A center leaning source frames the experiment as a biological advancement for synthetic biology and medicine. In contrast, a center left leaning source frames the discovery as a dual use dilemma, questioning whether such technology serves as a lifesaver or a potential bioweapon.