Barcelona /Washington, May 13 — Scientists have successfully used generative artificial intelligence to design synthetic DNA that can control gene expression in healthy mammalian cells, marking a major breakthrough in the field of synthetic biology.
The research, led by a team at the Centre for Genomic Regulation (CRG) in Barcelona, demonstrates for the first time that AI can create artificial gene-controlling elements, known as enhancers, capable of selectively switching genes on or off in specific cell types.
The team developed a language model for biology, enabling AI to generate DNA sequences of around 250 base pairs tailored to regulate gene activity in particular types of cells. When introduced into mouse blood cells, the synthetic sequences fused randomly with the genome and effectively triggered the expression of a fluorescent protein gene — without disrupting other gene expression patterns.
“The potential applications are vast. It’s like writing software, but for biology,” said Dr. Robert Fromel, the study’s first author. “This gives us new ways to instruct cells on how to grow and behave with unprecedented accuracy.”
The AI system was trained to understand the “grammar” of enhancer sequences — the short, non-coding segments of DNA that regulate gene activity — and to design new ones capable of operating in precise biological contexts. For instance, the model can be prompted to activate a gene in stem cells destined to become red blood cells, but not in cells that form platelets.
Although AI has previously been used in protein design, this study extends its use into controlling gene expression — a process at the root of many human diseases. Faulty gene activity in specific cell types often cannot be treated effectively with traditional protein-based drugs.
By engineering highly selective synthetic enhancers, researchers hope to develop new gene therapies that avoid unintended activation in healthy tissues, thus minimizing side effects.
“To create a language model for biology, you need to understand the language cells speak,” said Dr. Lars Velten, the study’s senior author. “We set out to decipher these grammar rules so we could create entirely new sentences — and now, we’ve shown that it works.”
The findings could help gene therapy developers fine-tune treatments and provide more targeted control over gene activity, offering promising avenues for future personalized medicine.
