AI-Powered Breakthrough Uncovers Hidden Genetic Variants Behind Common Diseases.

A collaborative team from the Children's Hospital of Philadelphia (CHOP) and the Perelman School of Medicine at the University of Pennsylvania has utilized a deep learning algorithm to uncover genetic variants in noncoding DNA that may elevate the risk for various common diseases. This innovative approach provides a significant step forward in interpreting the vast, uncharted regions of the human genome that do not code for proteins.

Although only a small portion of our genome encodes proteins, more than 98% consists of noncoding regions. These areas, long considered genomic “dark matter,” often regulate when and how genes are expressed. However, studying them has been particularly challenging due to their complex and poorly understood regulatory roles. Genome-wide association studies (GWAS) have previously flagged some of these regions as relevant to disease, but identifying the precise variant responsible among many has remained elusive.

The new study addresses this challenge using a technique that focuses on transcription factor binding—key regulatory proteins that control gene expression by attaching to specific DNA motifs. These binding sites typically occur in "open" sections of the genome. When transcription factors bind to DNA, they leave behind a unique "footprint" in experimental data, which helps pinpoint their exact binding location.

To identify these footprints, the researchers employed ATAC-seq, a genomic sequencing technique that detects accessible, or open, chromatin. They then applied a deep learning-based method called PRINT to analyze the resulting data. Working with liver tissue samples from 170 individuals, the team identified 809 footprint quantitative trait loci (footprint QTLs)—regions where DNA-protein interactions vary depending on the genetic variant present.

“This situation is comparable to a police lineup,” explained senior author Dr. Struan F.A. Grant, Director of the Center for Spatial and Functional Genomics at CHOP. “You’re looking at multiple suspects, but with this method, we can identify the exact culprit by detecting the footprint it leaves behind.”

With this foundational success, the researchers intend to expand their method to analyze other tissue types. The goal is to determine which noncoding variants are actively driving disease processes across different organs. According to Max Dudek, the study's first author and a PhD student at Penn Medicine and CHOP, larger datasets in future studies may pave the way for identifying causative variants that could eventually inform new therapeutic strategies.

Source: https://www.sciencedaily.com/releases/2025/04/250417145017.html

This is non-financial/medical advice and made using AI so could be wrong.