Every human cell depends on tens of thousands of genes being switched on at the right time and in the right amount. At the heart of this调控 lies a short DNA sequence known as the "initiator" — the precise point where a gene's information first begins to be converted into functional products like enzymes, hormones, and proteins.
Now, researchers in Professor James T. Kadonaga's laboratory at the University of California San Diego have decoded the initiator's DNA signature using machine learning. The work, published July 31 in Genes & Development, combined high-throughput sequencing of roughly 500,000 different versions of the initiator with AI models that learned the characteristic sequence pattern.
Once the signature was identified, the team searched the human genome and found that approximately 60% of human genes contain an initiator. The model provides, for the first time, strong predictions of whether a gene has an initiator — and crucially, can predict how mutations in this region might contribute to disease.
"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," said Kadonaga. He envisions that within the six billion bases of DNA in each of our cells, there exists a gene expression code specifying when, where, and to what extent each gene should be activated. "The new AI model for the initiator is a small but important part of this gene expression code."
The findings could help scientists create synthetic promoters — DNA sequences designed to switch genes on and off with specific customized functions — and could ultimately enable researchers to predict how different genetic variants affect individuals differently, a key step toward precision medicine.
