Researchers at the Centre for Genomic Regulation have created an artificial intelligence (AI) tool that designs new DNA sequences. These sequences, called synthetic enhancers, control gene expression, which is how cells turn genes on or off to perform tasks.
The AI makes DNA fragments with specific instructions, like turning on a gene in certain cells but not others. The study, published in Cell, shows these AI-designed DNA pieces work in healthy mouse cells.
The AI predicts the right combination of DNA letters, known as A, T, C, and G, to achieve desired gene activity. Scientists then build these 250-letter DNA fragments in a lab and insert them into a virus. The virus delivers the DNA into cells, where it merges with the cell’s genome. In tests, the researchers used the AI to design DNA that activated a gene for a glowing protein in some mouse blood cells without changing other gene patterns. The experiments worked as the AI predicted.
New possibilities for biology
This AI tool acts like a programmer for cells, offering precise control over gene activity. It could improve gene therapy. The tool designs enhancers, small DNA pieces that act like switches to control genes. Unlike natural enhancers, these AI-made ones can target specific cells, reducing side effects in healthy cells.
Until now, scientists mostly used AI to design proteins, like enzymes, which are molecules that speed up chemical reactions. However, many diseases come from incorrect gene expression in specific cells, and proteins may not always fix these issues. The AI tool creates enhancers that nature never made, offering new ways to fine-tune genes.
To build the AI, researchers ran thousands of experiments with healthy blood cells. They studied transcription factors, proteins that help control genes, and tested over 64,000 synthetic enhancers. These experiments showed how enhancers act like dials or switches, turning gene activity up, down, or off. Some combinations of transcription factors surprisingly shut genes off, a discovery called negative synergy.
The data helped the AI learn to design new enhancers with precise effects. This study proves the technology works, but researchers say much more exploration is needed.