In a discovery that feels straight out of science fiction, researchers in Barcelona have used artificial intelligence to create synthetic DNA that can control gene activity in healthy mammalian cells.
This is the first time this level of precision has been demonstrated outside of cancer models, and the opportunities it creates for medicine are significant.
Instead of rewriting the genes themselves, scientists at the Centre for Genomic Regulation have trained an AI to craft the genetic instructions that guide how cells behave. These short strands of DNA, known as enhancers, work like switches.
The AI can now create them from scratch to flip genes on in certain cells while keeping them silent in others.
Dr. Robert Frömel called the breakthrough “like writing software but for biology,” adding that it lets scientists guide how cells develop and behave “with un-precedented accuracy.”
The team spent five years building a massive library of over 64,000 synthetic DNA sequences, each one a tiny genetic switch.
They tested the sequences in healthy mouse blood stem cells, tracking how the sequences behaved as those cells developed into different blood types like red blood cells and platelets.
What they discovered has reshaped how scientists think about gene regulation. It turns out that transcription factors, the proteins responsible for turning genes on or off, do not always work alone.
In many cases, they interact in unexpected ways. Some combinations that typically activate genes can, when paired together, actually silence them.
The researchers call this “negative synergy,” and it appears to play a crucial role in how cells develop their specialized roles. This matters because most genetic therapies today still lack precision. They often get the job done, but can affect healthy cells along the way. It’s like trying to fix a single wire with a hammer. These AI-designed enhancers change that. They let scientists fine-tune gene activity in just the right cells, without disturbing the ones that don’t need it.
The team went a step further. They trained an AI model on their data, then asked it to design new DNA with specific goals, like turning on a gene in red blood cells but not in immune cells.
When tested in the lab, many of the AI’s designs worked just as predicted.
The impact could go far beyond blood. This approach could one day help build gene therapies for conditions like muscular dystrophy, diabetes or neurological disorders, where the hardest part is not fixing the gene but making sure the message gets to the right cells.
It is still early, but this marks a major shift in how we work with the genome.
Instead of relying solely on what evolution has given us, scientists can now design entirely new regulatory sequences, ones that have never existed in nature.
The research signals a turning point where biology meets engineering, and where imagination is beginning to shape what is scientifically possible.
This is more than a passing breakthrough. It is a moonshot for medicine, constructed carefully and precisely, one synthetic sequence at a time.
