A team at the Arc Institute led by Brian Hie has demonstrated the physical synthesis of functional biological agents using the Evo 1 and Evo 2 genomic foundation models. The researchers generated full bacteriophage genomes based on the architecture of the ΦX174 virus, moving computational designs from simulation straight into wet-lab experiments.
Experimental validation confirmed that the synthesized phages were viable. Furthermore, the generated pool of artificial phages successfully infected three strains of Escherichia coli that were naturally resistant to the original wild-type ΦX174 virus. The model did not merely reproduce a known sequence; it generated functional variations capable of overcoming bacterial defenses in laboratory tests.
For the R&D sector and deeptech investors, this milestone marks a fundamental engineering shift. Biological AI is moving beyond purely predictive tasks—such as 3D protein folding—to become a direct synthesis engine. This opens the door to the precision engineering of autonomous microbiological agents and targeted antimicrobial therapies tailored to specific resistance profiles.