Particle.news
Download on the App Store

AI System Designs Entire Viral Genomes as Lab-Built Phage Kills E. coli

A new high-accuracy DNA assembly method could speed construction of such designs, prompting calls for careful governance.

Overview

  • Stanford researchers used an AI model called Evo2 to generate 285 candidate bacteriophage genomes from scratch and screened them in petri-dish tests.
  • Sixteen AI-designed viruses showed activity against E. coli, and a mixture of those 16 overcame highly resistant bacterial strains in laboratory assays.
  • One minimal design, Evo-Φ2147, comprising 11 genes across about 5,386 base pairs, was synthesized and demonstrated bactericidal effects in vitro.
  • Caltech scientists reported a DNA construction tool named Sidewinder in Nature that claims much higher accuracy for assembling long genetic sequences.
  • Developers and advisers cite potential medical applications and say safeguards were used, including not training on human-pathogenic viruses, while urging clearer oversight.