Researchers have successfully used artificial intelligence to design and synthesize fully functional viral genomes from scratch, creating artificial bacteriophages capable of destroying antibiotic-resistant bacteria. Published without peer review on bioRxiv, the breakthrough demonstrates computer-assisted genome writing.
For the first time, scientists have bridged the gap between computational code and biology by deploying generative artificial intelligence to write coherent whole-genome viral sequences. Rather than simply editing existing organisms, the team of researchers generated new genetic instructions from the ground up. These computer-designed blueprints were then synthesized into physical bacteriophages—viruses that hunt and destroy bacteria—specifically targeting strains of E. coli, including those resistant to standard antibiotic treatments.
The achievement relies on advanced generative models to tackle a biological frontier that has long resisted automation. While AI tools have previously excelled at generating short DNA strings, individual proteins, or complex molecular structures, orchestrating an entire genome requires managing intricate interactions across thousands of base pairs.
How Evo Models Built an Artificial Virus
The scientists utilized Evo 1 and Evo 2 AI models, which are engineered to analyze and generate DNA, RNA, and protein sequences. These systems were originally pre-trained on more than two million phage genomes.
The process required a precise design template to guide the algorithms. Researchers selected ΦX174, a simple single-stranded DNA virus featuring 5,386 nucleotide bases distributed across 11 genes. This compact viral template provides all the necessary genetic machinery to infect a host and replicate successfully. Using supervised learning techniques, the models were directed to generate viral genomes similar to ΦX174 that could selectively target and destroy stubborn bacterial pathogens.
The models generated thousands of candidate sequences. After rigorous screening, the team winnowed the output down to 302 synthetic bacteriophages that possessed the correct functional capabilities. Laboratory testing confirmed that these artificial constructs could successfully infect and kill their designated bacterial hosts.
Expert Reactions and the Horizon of AI-Assisted Genome Writing
The scientific community has greeted the preprint—which was uploaded to the bioRxiv server—as a moment for computational biology. Because viruses are technically not living entities, researchers emphasize that creating fully autonomous cellular life remains a formidable hurdle. Even so, the implications for genetic engineering are profound.
This is the first time artificial intelligence systems have been able to write coherent sequences on a whole-genome scale. Brian Hie, computational biologist at Stanford University in California
Prof Patrick Cai, chair of synthetic genomics at the Manchester Institute of Biotechnology, stated that the significance extends far beyond phages, suggesting that genome language models are beginning to learn the design principles encoded by evolution and thereby opening the door to AI-assisted genome writing.
Additional commentary emphasized the nature of the work. Prof Marc Güell, from the synthetic biology lab at Pompeu Fabra University in Spain, described the development as a very significant turning point that invites researchers to imagine novel solutions for immunotherapy, genetic disorders, and resistant infections.
Scaling from Phages to Complex Organisms
Despite the excitement, the distance between designing a small viral phage and engineering complex life is vast. The synthetic phage genome spans roughly 5,400 base pairs. By comparison, the smallest known genome of a living cell contains approximately 500,000 base pairs, while the human genome comprises three billion.
Researchers involved in the project acknowledge that navigating this scale will require immense effort. While computational pioneers suggest that attempting simpler organisms is definitely interested in working towards, experimental biologists caution that considerable hurdles remain before building a complete living organism becomes feasible.
A lot of experimental breakthroughs must happen before designing a whole living organism becomes possible. Samuel King
As the scientific community reviews the findings and moves toward formal peer review, the primary objective centers on therapeutic applications. Investigators hope engineered phages will eventually complement current medical treatments and bolster defenses against dangerous pathogens.
