Precise activation of tens of thousands of genes is critical for healthy development and growth. Specialized segments of DNA orchestrate the production of enzymes, hormones, proteins, and other components essential for cell structure and function. When genes are not correctly activated, cells can stop functioning or contribute to various disorders, including cancer. Now, researchers at the University of California San Diego have applied artificial intelligence to decode a key DNA element involved in this process.

The element, known as the "initiator," marks the location where information coded in a gene begins to be converted, or expressed, into a functional product. In a study led by graduate student researcher Torrey Rhyne-Carrigg, scientists in the laboratory of Professor James T. Kadonaga used high-throughput DNA sequencing to measure gene expression activity across approximately 500,000 different versions of the initiator. They then trained a machine learning system, a form of AI, to identify the characteristic DNA pattern associated with the initiator. Once the model 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.

The findings could help researchers predict how DNA mutations affecting the initiator may alter gene activity and contribute to disorders. 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 purposes.

More broadly, the research demonstrates how laboratory experiments and artificial intelligence can be combined to uncover information embedded in human DNA. Kadonaga noted that each human cell contains six billion bases of DNA, and within that sequence lies a gene expression code that specifies when, where, and to what extent each gene should be turned on or off. He expressed hope that AI models of the entire gene expression code could be expanded in the not-too-distant future, allowing prediction of the activity of different gene variants in different people.

The study was published in Genes & Development with the title "Machine learning analysis of the human initiator region reveals key features of different types of core promoters" (Torrey E. Rhyne-Carrigg et al., 2026, DOI: 10.1101/gad.353623.125).