Creating viruses to fight bacteria: the idea seems paradoxical. However, some viruses only infect bacteria. These are the bacteriophages. Researchers at Stanford and the Arc Institute are using artificial intelligence models called Evo to propose new genomes.
The computer proposes sequences; the researchers then carry out the experiments. In their study published in August two thousand twenty-six, sixteen phages are viable in the laboratory. An IT proposal is therefore not enough: its operation must be established in the real world.
A mixture of these phages also overcomes the resistance of Escherichia coli strains to the reference phage. This result relates to these bacteria and conditions. It does not demonstrate that all antibiotic-resistant microbes could be eliminated in this way.
The challenge is to develop other avenues against bacterial infections. But no cured patient is demonstrated by this study, nor a strengthening of existing antibiotics. The advance combines the proposal by AI and human verification; the transition to treatment remains to be established.
What we know
Laboratory results on bacteria. Resistance overcome is that to the reference phage; this study demonstrates neither clinical treatment nor reinforcement of existing antibiotics. The illustrations are educational and do not constitute images of the real essay.
To go to the source
- Science — Generative design of bacteriophages with genome language models ↗
- Stanford — AI designs a novel E. coli killer ↗
References consulted on 16 October 2026.
Signaler une erreur ou apporter une précision
