In a scientific breakthrough, researchers have used artificial intelligence to design synthetic DNA that can precisely control gene activity in living mammalian cells—paving the way for next-generation gene therapies.
Key Points at a Glance
- AI was used to generate novel DNA sequences that regulate gene expression in specific cell types.
- These synthetic enhancers were successfully tested in healthy mammalian cells.
- The approach enables targeted, cell-specific gene therapies with fewer side effects.
- This marks a milestone in the emerging field of generative biology.
In a pioneering study that blends biology with cutting-edge artificial intelligence, scientists at the Centre for Genomic Regulation (CRG) in Barcelona have demonstrated that AI can design entirely new DNA sequences capable of controlling genes inside healthy mammalian cells. This first-of-its-kind achievement signals a profound advancement in our ability to manipulate genetic activity—and opens the door to a new era of precision medicine.
Every cell in the human body contains the same DNA blueprint, yet different genes are turned on or off depending on the cell type and function. This control is governed by DNA regions called enhancers—short sequences that act like genetic switches. Designing synthetic enhancers that can control gene expression with high accuracy has long been a goal of synthetic biology, but until now, the process relied heavily on trial and error and natural templates.
The CRG team, however, flipped the script by developing a deep learning model that can generate entirely novel DNA sequences tailored to drive gene expression in specific cell types. Using massive datasets of known DNA enhancers, the AI learned the statistical rules and patterns that determine how these genetic switches behave. Then, it created new enhancer sequences that do not exist in nature—but work as if they do.
To validate the AI’s predictions, the researchers synthesized several of these custom-designed enhancer sequences—each about 250 base pairs long—and inserted them into mouse hematopoietic (blood) cells. To test whether the enhancers were functional, they linked them to a fluorescent reporter gene. The results were stunning: the cells lit up exactly as intended, proving that the AI-generated sequences could turn genes on with remarkable precision.
Why is this important? Because in the field of gene therapy, precision is everything. Traditional approaches often deliver therapeutic genes into the body using viral vectors, but these methods can lack control over where and how strongly the genes are activated. This imprecision can lead to unintended effects or reduced treatment efficacy. With synthetic, AI-designed enhancers, scientists can build genetic circuits that activate only in desired cells—reducing collateral damage and improving outcomes.
This concept is known as “cell-type specificity,” and it’s one of the holy grails of genetic medicine. For example, a treatment for a liver disorder should ideally act only in liver cells—not in muscle, brain, or blood. By engineering enhancers that are active only in target cell types, researchers can ensure that therapies are both safer and more effective.
Moreover, this work ushers in a new field that the CRG researchers call “generative biology.” Just as generative AI models can produce realistic images, text, or music, these biological models can now “imagine” and create DNA sequences with desired functions. Instead of editing or copying what evolution has already made, we can now instruct AI to invent entirely new genetic elements—with therapeutic potential.
The implications are enormous. Future applications could include personalized treatments where enhancers are tailored to a patient’s unique cellular profile, or synthetic organs built from the ground up with AI-designed control systems. The tools developed in this study may also help scientists better understand the logic of gene regulation, a complex puzzle that remains one of the great challenges in modern biology.
Still, the technology is in its early days. More work is needed to refine the models, ensure long-term safety, and expand their use to other types of cells and organisms. However, the proof of concept has been established. AI can not only read and interpret genetic code—it can write it.
In a world increasingly shaped by both biotechnology and machine learning, this union of disciplines could revolutionize how we treat disease, design organisms, and understand the fundamental mechanics of life.
Source: Center for Genomic Regulation
