AI Designs Living Viruses From Scratch. Sixteen Actually Work. Nature Is No Longer The Only Author.
Researchers deployed AI models called Evo 1 and Evo 2, trained on massive genomic sequence collections, to design entirely new bacteriophage genomes. They synthesized and tested hundreds of AI-generated designs experimentally. Sixteen produced functional bacteriophages capable of infecting and killing Escherichia coli.
This demonstrates generative biology, the principle that AI can move beyond pattern recognition into functional de novo design of living systems. The mechanism is sequence-to-function prediction at scale. The reader should understand that AI is transitioning from a tool that reads nature to one that authors it.
The researchers behind the Evo 1 and Evo 2 models, who used them to generate novel bacteriophage genomes rather than search for existing ones in nature.
- Open a free account on ESM Atlas (esmatlas.com), a consumer-accessible protein folding predictor, and paste in a random amino acid sequence of 50 characters. Expected outcome: You get a predicted 3D protein structure, demonstrating how AI maps sequence to function.
- Try the same sequence on AlphaFold Server (alphafoldserver.com), Google's free consumer tool. Expected outcome: A structural prediction with confidence scores, showing how different models interpret the same biological input.
- Compare the two outputs side by side. Expected outcome: You observe that both models infer structure from sequence, the same fundamental principle that allowed Evo to design functional viruses from genomic patterns.