Research2026-08-06
Stanford researchers used Evo 2, a generative AI model that writes DNA sequences, to design complete genomes for bacteria-killing viruses, then built and tested them against E. coli. Sixteen of the roughly 300 designs worked well, and a mixture of them overcame bacterial resistance in lab tests. The work is a proof of concept published in Science, not a treatment, and the model is available free to download.
What changed
Evo 2's genome designs had largely stayed computational, untested in the lab.
What it unlocks
Designing complete viral genomes with an AI model and validating them in the lab, using a freely downloadable tool.
- ~300 phages synthesized and tested
- 16 exceptional E. coli killers
- ΦX174 genome under 6,000 base pairs
What you need to act on it
- a wet lab with DNA synthesis capability
- compute to run Evo 2
Sources