AI Synthesises 302 Novel Viral Genomes

An artificial intelligence model has successfully generated 302 novel viral genomes. This development marks the first instance of an AI creating full, functional viral genetic sequences.

Scientists at Stanford University and the ARC Institute conducted this research. Their AI model, named Evo, received training on a dataset of known bacteriophage genomes. Evo subsequently learned to differentiate between viral and non-viral genetic material. The model's objective was to design genetic sequences that would express functional viral proteins within a bacterial host.

Evo generated a total of 1500 candidate viral genomes. The research team synthesised 200 of these candidates for experimental validation. Of the synthesised genomes, 100 were observed to produce functional viruses. An additional 202 genomes were also identified as potentially viable.

These experiments confirmed that Evo could produce synthetic viruses with similar infectious properties to natural viruses. Some of the AI-designed genomes exhibited infectivity comparable to naturally occurring phages. The project's findings were published in a peer-reviewed journal.

This demonstration of AI-driven viral design capability raises pertinent questions for biosecurity. The ability to create novel, functional biological agents presents new challenges. Researchers emphasised the necessity of proactive safety measures alongside such technological advancements. The work highlights the growing need for responsible innovation frameworks in AI and synthetic biology.

What new considerations arise for regulating AI systems capable of de novo biological design?