AI designs synthetic viruses to combat bacteria

Scientists are leveraging artificial intelligence to engineer entirely new viruses capable of combating drug-resistant bacteria, marking a significant step in synthetic biology. Researchers, led by Stanford University, have successfully designed and synthesized complete, functional genomes for bacteriophages, viruses that infect bacteria.

“

were different to any fago natural known, exhibiting mutations new, genes and elements reguladores divergentes

Using the bacteriophage FX174, which targets E. coli, as a design model, the AI combined genomic language models, computational biology, and experimental screening. The AI generated thousands of new, similar genomes, from which scientists chemically synthesized and tested nearly 300. This process yielded 16 viable phages that were distinct from any known natural phages, featuring novel mutations and regulatory elements.

Crucially, some of these AI-designed phages demonstrated effectiveness against E. coli strains that had developed resistance to similar natural phages. This breakthrough suggests a potential new avenue for developing phage therapies to combat rapidly evolving bacterial pathogens. In comparative experiments, some AI-generated phages even outperformed natural ones in replication speed and competitiveness within the same cell culture.

“

opens a way towards therapies with phages generated by AI against bacterial pathogens in rapid evolution

However, the ability to design and synthesize functional viral genomes using AI also brings significant biosecurity and biosafety considerations to the forefront. The study's authors emphasize the need for a robust debate on governance, policies, and mitigation strategies to prevent misuse. They recommend that research groups involved in designing complete genomes consult with security and biosecurity professionals throughout project lifecycles, suggesting that existing safety frameworks can be adapted and that risk mitigation can be enhanced by excluding sensitive viral sequences from AI training data.

“

important considerations in terms of biosecurity, biocontention and biological safety