Healthy growth and development rely on tens of thousands of genes being activated at the right time and in the right cells. Specialized DNA regions regulate this process by controlling the production of enzymes, hormones, proteins, and other molecules essential for cell function. When gene activation goes wrong, cells can malfunction and contribute to diseases such as cancer.
To better understand the DNA sequences that regulate gene activity, researchers in the laboratory of a UC San Diego professor focused on key genetic elements called “initiators.” These DNA sequences identify where the information in a gene begins to be transcribed and converted into a functional product.
Artificial intelligence decodes DNA initiator sequences
In the new study, led by graduate student researcher Tori Lyne Carrig, the team used high-throughput DNA sequencing to measure gene expression across approximately 500,000 different versions of an initiator sequence.
The researchers used the results to train a machine learning model—an artificial intelligence system designed to recognize patterns in biological data. The model identified distinctive DNA features associated with initiators. After decoding this sequence signature, the team searched human genes and discovered that approximately 60% contain an initiator.
Professor Kadonaga of the Department of Biological Sciences and Molecular Biology at the University of California San Diego said, “For the first time, we found that AI models can strongly predict the presence or absence of an initiator in a human gene, and we were able to decipher the DNA sequence pattern of the initiator.”
Using AI to predict the effects of DNA mutations
This discovery may help scientists predict how mutations in initiator sequences alter gene activity and contribute to disease. The study’s data and AI models could also support the development of synthetic promoters with customized functions. These engineered DNA sequences can be designed to switch genes on or off with greater precision.
More broadly, the research demonstrates how laboratory experiments and artificial intelligence can work together to uncover the regulatory information encoded in human DNA.
“More broadly, this study is a step forward in combining laboratory experiments and AI to decipher the information embedded in human DNA sequences,” Kadonaga said. “Ultimately, the 6 billion DNA bases in each of our cells contain a gene expression code that determines when, where, and how strongly each gene is activated or silenced. If we could develop an AI model representing the entire gene expression code, we might predict the activity of different gene variants in different people. The new AI model for initiators is a small but important part of this code, and I am optimistic that we will be able to expand it to represent human gene expression in the not-too-distant future.”
Source: www.sciencedaily.com