Healthy growth and development depend on tens of thousands of genes being switched on at the right time and in the right place. Specific regions of DNA help coordinate this process, guiding the production of enzymes, hormones, proteins, and other molecules that cells need to function properly. When gene activation goes wrong, cells can malfunction and contribute to diseases, including cancer.
To better understand the DNA sequences that control this process, researchers in the laboratory of University of California San Diego Professor James T. Kadonaga focused on an important DNA element known as the 'initiator.' The initiator marks the location where the information encoded in a gene begins to be converted, or expressed, into a functional product.
AI Decodes the Initiator Sequence
In the new study, led by graduate student researcher Torrey Rhyne-Carrigg, the team used high-throughput DNA sequencing to measure gene expression activity across approximately 500,000 different versions of the initiator.
The researchers then used those results to train a machine learning system, a form of artificial intelligence, to identify the characteristic DNA pattern associated with the initiator. Once the model had decoded that signature, the team searched human genes for the sequence and found that roughly 60% contain the initiator.
"These AI models were found to provide, for the first time, strong predictions of the presence or absence of the initiator in human genes, and were thus able to decode the DNA base sequence pattern of the initiator," said Kadonaga, a professor in the UC San Diego Department of Molecular Biology, School of Biological Sciences.
Implications for Predicting Mutation Effects
The breakthrough could help predict the effects of harmful mutations, which may disrupt the initiator and lead to misregulated gene expression. By understanding the exact DNA signature that controls gene activation, researchers could identify disease-causing variants more accurately and potentially develop targeted therapies.
Looking ahead, the team plans to expand this approach to decode other regulatory elements in the human genome, aiming to build a comprehensive map of the genetic instructions that control gene activity throughout the body. In 2026, with the rapid advancement of AI in genomics, such models are becoming increasingly powerful, enabling scientists to unravel complex biological codes at an unprecedented pace.
The findings were published on August 23, 2026, and represent a significant step toward deciphering the full regulatory language of human DNA. As AI continues to evolve, it promises to unlock even more secrets hidden within our genetic blueprint, offering new possibilities for personalized medicine and disease prevention.