Prime Highlights
- Stanford University and the Arc Institute have developed the world’s first fully AI-generated viral genome, capable of destroying disease-causing bacteria.
- The breakthrough shows AI’s potential to accelerate alternatives to antibiotics, gene therapy, and crop defense.
Key Facts
- Using the AI system Evo, researchers designed thousands of viral genomes, with 16 successfully infecting and destroying E. coli, including strains resistant to natural viruses.
- The study, currently in preprint on bioRxiv, is the first demonstration of AI creating entire functioning viral genomes, though experts warn of risks if applied to harmful pathogens.
Background
For the first time, researchers have used artificial intelligence to create viruses that destroy disease-causing bacteria. The research, led by Stanford University and the Arc Institute, produced the world’s first fully AI-generated viral genome, a major step in biology.
The team specialized in bacteriophages, which are viruses that infect bacteria. The researchers used an AI system, called Evo, which has been trained on millions of viral genomes, to design thousands of genome designs. Among 302 designs synthesized, 16 of them were able to infect and destroy Escherichia coli successfully in laboratory tests, including some strains that were not killed by natural viruses.
The study used phiX174, a small virus with 11 genes, as a test case. Some AI-designed viruses included many new mutations. One design even used a DNA-packaging protein from a distant virus, something human scientists had not been able to do before.
This work demonstrates that AI is capable of creating entire functioning genomes, which may accelerate the creation of alternatives to antibiotics, gene therapy, and crop defense. According to scientists, AI might assist in the creation of viruses with specific bacterial adaptation in haste compared to conventional methods.
The team emphasized that the viruses were tested on safe lab strains. Experts also caution against the possible risks in case such methods were applied to hazardous pathogens.
The study is a preprint on bioRxiv and will go through peer review to confirm the results, but it marks an important milestone for AI in biotechnology.
