AI designs world’s first functioning virus genome — a new frontier in synthetic biology

The AI-generated virus, capable of infecting and killing bacteria, represents a milestone achievement: while scientists have previously used AI to design proteins and small genetic systems, constructing a complete genome — a self-contained set of instructions that governs an organism’s life processes — had never before been achieved.

A new phase in genome design

“Genome design requires orchestrating multiple interacting genes and regulatory elements while maintaining a balance that enables replication, host specificity, and evolutionary fitness,” the research team explained in a post on arcinstitute.org. “This increase in complexity introduces new constraints and failure modes that do not arise when only designing a single protein or a two-component system.”

The team selected the bacteriophage ΦX174 (pronounced “phi-X-174”) as their test subject — a small virus that infects E. coli bacteria. Its genome contains 5,386 DNA letters and 11 overlapping genes, making it both compact and challenging to model. Notably, ΦX174 was the first genome ever sequenced (in 1977) and the first to be chemically synthesized (in 2003). Now, it has become the first virus designed entirely by AI.

Training AI to “speak” the language of DNA

The scientists trained an advanced genomic language model, called Evo, on thousands of related viral genomes, effectively teaching it the genetic “grammar” of ΦX174. Using text-based prompts, Evo generated thousands of novel genome designs.

To ensure biological feasibility, researchers developed custom software to verify that each AI-designed genome contained the essential genes and proteins required to infect E. coli. They then synthesized hundreds of these genomes in the lab and tested them in bacterial cultures.

The results were remarkable: 16 new, fully functional viruses emerged. Each carried dozens to hundreds of genetic mutations that had never been seen in nature.

One particularly innovative design borrowed a DNA-packaging protein from a distant viral relative — a modification that had repeatedly failed in human-led experiments. Using cryo-electron microscopy, scientists confirmed that the hybrid protein functioned perfectly within the viral shell.

AI outperforms natural evolution

In controlled tests, some AI-designed viruses even outperformed their natural counterparts. When E. coli bacteria evolved resistance to natural ΦX174 strains, several AI-created viruses breached those defenses within days.

Across 302 AI-generated candidates, 16 proved viable, and the designs featured 392 previously unknown genetic mutations — combinations that had eluded human genetic engineers for decades.

A new era in biotechnology

The achievement represents the next phase in humanity’s ability to manipulate life: moving from reading DNA (sequencing), to writing it (synthesis), and now to designing it using AI.

Experts say the experiment showcases AI’s transformative potential in biotechnology — not merely to speed up discovery, but to explore vast genetic possibilities beyond the limits of natural evolution.

“This work demonstrates how AI can go from analyzing data to engineering life itself,” said one computational biologist not involved in the study. “It’s an entirely new level of scientific creativity.”

As artificial intelligence continues to advance across fields from education to engineering, this breakthrough signals the dawn of AI-driven biology — where life can be intelligently designed, optimized, and understood in ways once thought impossible.