AI Designs First Complete, Viable Genomes for Viruses to Kill E. Coli
Stanford researchers have successfully used genomic foundation models to design 16 novel, fully functional bacteriophages from scratch that can infect and kill E. coli. This marks the first time artificial intelligence has generated complete, viable genomes for living viruses. This breakthrough demonstrates AI's capability to design biological systems beyond what exists in nature, paving the way for custom-designed phage therapies to combat antibiotic-resistant bacterial infections. It represents a major milestone in synthetic biology and AI-driven drug discovery. The researchers synthesized 302 AI-generated designs in the lab, resulting in 16 viable bacteriophages with genomes of approximately 5,400 base pairs. The underlying AI models, Evo 1 and Evo 2, operate similarly to large language models but predict genomic sequences instead of text.
## BACKGROUND
Bacteriophages, or phages, are viruses that selectively infect and destroy specific bacteria without harming human cells. Phage therapy is an alternative to traditional antibiotics, which is increasingly crucial as antibiotic-resistant "superbugs" become a global health threat. Evo is a genomic foundation model trained on prokaryotic DNA to predict and design biological sequences at single-nucleotide resolution.