AI-generated bacteriophage genomes that replicate and kill bacteria
Genome language models wrote complete ΦX174-like bacteriophage genomes from scratch; of 285 designs that could be built, 16 produced viable infectious phages, several with faster lysis than the natural virus.
- Model
- Evo 1 and Evo 2
- Field
- Biology
- Date
- 2025-09-17
- Human collaborators
- Samuel King, Brian Hie
Sources
Original work
Announcement
Media coverage
Independent commentary
What was found
The team fine-tuned Evo 1 and Evo 2 on a cleaned set of nearly 15,000 Microviridae genomes, generated 302 candidate genomes with ΦX174-like architecture, chemically synthesised the 285 that could be assembled, and tested them for plaque formation in E. coli. Sixteen were viable, with substantial sequence divergence from the template. Several generated phages outcompeted wild-type ΦX174 in growth and lysis kinetics, and a cocktail of them cleared three ΦX174-resistant E. coli strains. The authors frame it as the first generative design of a complete functional genome.
Novelty check
Synthetic phage genomes assembled from natural sequence date to Venter's 2003 ΦX174 reconstruction, and generative design of individual proteins was well established. Generating an entire genome that yields a working organism had not been demonstrated. The claim is bounded to bacteriophages, which infect bacteria only.
Caveats and known objections
Peer review does not replicate: no independent group has rebuilt these genomes, and the 16 viable designs remain the authors' own result. The designs are variants within one small, exceptionally well-characterised template (ΦX174, 5,386 bp), not free-form genomes, and 269 of the 285 that were built did not work. On biosafety, the authors excluded eukaryotic and human-infecting viruses from training and confine the work to phages; the result has nonetheless become a reference point in biosecurity debate, which is context this entry records rather than adjudicates. Science published a biosecurity perspective alongside the paper arguing for legally required screening of synthetic DNA orders. Note that some press coverage of the Science paper misattributes the work to OpenAI; the collaboration is Arc Institute and Stanford.
Nobody outside the lab has checked this yet.
Reading the primary source closely enough to say whether it supports the claim counts as a check, and you are credited on the entry.
Or on GitHub: submit a check challenge the grade send a correction or send a pull request
Flag this for triage
Signals order the review queue and nothing else. They are never published, and they never move a grade: that takes a citation.
History of this entry
- RegradedUpgraded from author verified to peer reviewed: the preprint was published in Science, alongside a biosecurity perspective calling for mandatory screening of synthetic DNA orders. source ↗
- AddedEntered the registry graded Peer reviewed and AI-led.
Entries are never deleted. A grade that does not hold up is downgraded on the record, with the reason beside it.
Graded peer reviewed for verification and ai-led for autonomy. What these mean.
Cite this entry
whataifound.org. (2025). AI-generated bacteriophage genomes that replicate and kill bacteria. whataifound.org: A Registry of AI Scientific and Mathematical Discoveries. https://whataifound.org/finding/2025-09-17-evo-phage-genomesBibTeX
@misc{whataifound-arcinstitute-2025-genomes,
title = {AI-generated bacteriophage genomes that replicate and kill bacteria},
author = {{whataifound.org}},
year = {2025},
howpublished = {whataifound.org: A Registry of AI Scientific and Mathematical Discoveries},
note = {Result by Arc Institute / Stanford University. Verification: Peer reviewed. Autonomy: AI-led.},
url = {https://whataifound.org/finding/2025-09-17-evo-phage-genomes}
}