Sixteen AI-designed viruses offer a new route against drug-resistant bacteria

For the first time, researchers from Stanford University, led by a team, have created 16 artificial viruses that do not naturally exist. These designer viruses, described in the journal Science, could potentially combat drug-resistant bacteria. Published alongside a Perspective piece by Thomas Inglesby and Moritz S. Hanke, the study demonstrates how artificial intelligence (AI) could revolutionize drug discovery by designing functional genomes based on natural virus templates.

The scientists employed specialized AI models trained on millions of DNA sequences, then refined these models on 15,000 viral genomes from the Microviridae family. This allowed the AI to grasp the genetic principles that govern functional viral genomes. From this digital design, the researchers synthesized 285 DNA candidates, which were then inserted into E. coli cells to produce 16 fully functional, reproducing viruses.

These 16 viruses successfully infected and destroyed targeted E. coli bacteria, and mixtures of these synthetic phages also killed bacterial strains resistant to the natural PhiX174 virus. The study’s authors emphasize that while AI-designed viruses hold promise for combating superbugs, clinical applications necessitate thorough testing.

Additionally, safety measures were taken to exclude human-infecting viruses from the AI’s training data. The researchers caution that as AI genome tools advance, biosecurity rules and oversight must keep pace.