Artificial intelligence has cracked a key piece of DNA's hidden genetic control system, identifying the DNA signature of a critical "on switch" that tells genes when to activate. Researchers at UC San Diego analyzed approximately 500,000 DNA sequences and trained a machine learning model to recognize the characteristic pattern of the "initiator," a DNA element that marks where genetic information begins to be converted into functional products. The model successfully identified this initiator in roughly 60% of human genes, offering scientists a new tool to understand how genetic activity goes wrong in disease.

What Is the Initiator and Why Does It Matter?

Every cell in your body contains about six billion DNA bases, and healthy growth depends on tens of thousands of genes being switched on at precisely the right time and place. The initiator is one of the DNA sequences that coordinates this process, guiding the production of enzymes, hormones, proteins, and other molecules cells need to function. When gene activation fails, cells can malfunction and contribute to diseases including cancer. Until now, scientists lacked a reliable way to identify and predict how changes to the initiator might affect gene activity.

The research, led by graduate student Torrey Rhyne-Carrigg in the laboratory of UC San Diego Professor James T. Kadonaga, represents a significant step forward in decoding the genetic instructions that control our biology. The team used high-throughput DNA sequencing to measure gene expression activity across hundreds of thousands of initiator variants, then fed those results into a machine learning system to identify the pattern.

How Can This Discovery Help Predict Disease Risk?

The findings open new possibilities for predicting how mutations affecting the initiator may alter gene activity and contribute to a range of disorders. By understanding the DNA signature of a functional initiator, researchers can now identify when mutations disrupt this critical switch, potentially causing disease. The study's data and AI models may also support the design of synthetic promoters, sequences that can switch genes on or off with functions tailored for specific research or therapeutic purposes.

"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 James T. Kadonaga, a professor in the UC San Diego Department of Molecular Biology, School of Biological Sciences.

James T. Kadonaga, Professor of Molecular Biology at UC San Diego

How AI and Lab Experiments Are Reshaping Genomics

This breakthrough demonstrates how combining laboratory experiments with artificial intelligence can unlock information hidden within human DNA. The approach is not limited to the initiator; it represents a broader strategy for decoding the entire gene expression code embedded in the six billion DNA bases in each cell. If scientists can expand these AI models to cover all the genetic switches and regulatory elements, they could eventually predict the activity of every gene variant in different people.

• Sample Size: The AI model was trained on approximately 500,000 different versions of the initiator DNA sequence to identify the characteristic pattern.
• Detection Rate: The model successfully identified the initiator in roughly 60% of human genes, providing the first reliable method for recognizing this critical genetic switch.
• Practical Applications: The findings could help predict how mutations affect gene activity, support the design of synthetic promoters for research, and contribute to understanding disease mechanisms.
• Future Scope: The initiator model is described as a small but important part of the broader gene expression code that controls when, where, and to what extent genes turn on or off.

"More globally, this work is a step forward in the combined use of laboratory experiments and AI to decipher the information that is embedded in the sequence of the DNA bases in humans. Ultimately, within the six billion bases of DNA in each of our cells, there is a gene expression code that specifies when, where and to what extent each of our genes should be turned on or off," noted Kadonaga.

James T. Kadonaga, Professor of Molecular Biology at UC San Diego

The research was published in the journal Genes in August 2026 and represents a meaningful advance in personalized medicine. By understanding how genetic switches work at the molecular level, researchers can better predict individual responses to mutations and potentially develop more targeted treatments for genetic disorders. Kadonaga expressed optimism that the team will expand its AI models of the human gene expression code in the near future, moving closer to a comprehensive understanding of how our genetic instructions control life itself.