Scientists have used large genome models to design new viruses that infect bacteria. This marks a major step forward in AI-driven biology, moving beyond protein design to whole-genome engineering.
How Large Genome Models Design New Viruses
Most AI work in biology has focused on designing proteins. That is because proteins do most of the work in living cells — they catalyze chemical reactions and provide structure. Designing a new protein means directly changing biochemistry, which can create new and useful functions.
But the genetic code sits between DNA and proteins. It was not obvious what a model trained on DNA could actually do. Despite that uncertainty, researchers built large genome models anyway. The results surprised them.
According to Nature, these models can now output DNA sequences that encode functional proteins in bacteria. They can also mimic the gene structures found in complex cells. Now, the same models have been used to output the genomes of viruses that infect bacteria.
From Protein Design to Full Genome Generation
The shift from proteins to genomes matters because it changes what AI can build. Protein design gives you one molecule at a time. Genome design gives you an entire biological system — in this case, a complete virus.
These "genome-language models" analyze thousands of DNA sequences to predict and build entirely new biological structures, as noted by Science Nature Page. The models learn the patterns in DNA the same way language models learn patterns in text. Then they generate new sequences that follow those patterns.
This is not science fiction. The models have already produced working genomes. The viruses they design are real, and they target bacteria.
What This Means for Biology and Medicine
The ability to design viruses that infect bacteria opens practical doors. Bacteriophages — viruses that attack bacteria — are already being studied as alternatives to antibiotics. AI-designed versions could be tailored to target specific resistant bacteria.
According to Professor Erwin Loh on LinkedIn, these AI-designed viruses target resistant bacteria. That is a direct and immediate use case for the technology.
The research is still early. But the direction is clear: AI can now design at the genome level, not just the protein level. That changes what is possible in synthetic biology.
Our Take: A Powerful Tool That Needs Careful Handling
To put it plainly, this is a significant achievement. Large genome models designing functional viruses is a leap beyond protein design. It shows that AI can grasp the full complexity of a genome, not just individual parts.
But this power cuts both ways. The same models that design viruses to fight bacteria could, in theory, be used to design harmful viruses. The technology is dual-use, and that reality needs to be part of the conversation from the start.
For now, the practical benefits are clear — AI-designed phages could help with antibiotic resistance, which is a growing global problem. But as this technology matures, researchers and regulators will need to think carefully about safeguards. The science is moving fast. The policy around it should not lag behind.
What matters most is that this work is transparent and published openly, so the scientific community can study it, verify it, and build on it responsibly.
