Machine Learning Reveals New Biological Variability in Gene Regulation

Researchers from University of Utah Health and collaborating institutions have developed FIRE, a tool that combines machine learning with advanced long-read DNA sequencing to analyze differential gene regulation between maternally and paternally inherited chromosomes. The findings were published in Cell and provide a sharper view of the largely noncoding regions of the genome, which represent about 98% of DNA and include many regulatory elements that control whether genes are turned on or off. FIRE builds on a method that identifies physically accessible regions of DNA, which can mark active genes or regulatory sequences. Because the tool sequences long stretches of individual DNA molecules, researchers can usually determine whether each stretch came from the mother’s or father’s copy based on small sequence differences. A machine learning component then evaluates billions of accessible regions across the genome to distinguish regulatory elements likely to have functional importance from regions that are merely accessible without biological significance. The approach was applied to a previously unresolved question in celiac disease genetics. Earlier studies had identified a DNA variant associated with disease risk, but it remained unclear whether that variant affected gene activity or was simply a linked marker. Using FIRE, the researchers found that the associated change sits in a regulatory switch near an immune gene also linked to celiac disease. The DNA change disrupts a binding site for gene-regulating proteins, and altering that single base in laboratory experiments changed the activity of the switch. This evidence suggests the variant is likely the causal genetic difference among many nearby changes. The team also used FIRE to identify tens of thousands of regulatory regions that had been previously overlooked. These regions are more likely than other parts of the genome to contain DNA variants associated with human disease, indicating they may play important but still poorly understood roles in health and disease. One of the study’s central discoveries concerns non-genetic biological variability. In many cases, differences in gene regulation between maternal and paternal copies could be explained by differences in DNA sequence. However, the researchers found that in roughly 40% of observed regulatory differences, the two copies were genetically identical at that location. Some of these regulatory differences were stable and passed on as cells divided, suggesting a layer of biological variation that is not determined solely by the genetic code. The authors say FIRE could help translate genetic associations into functional insights by linking DNA changes to measurable effects on gene regulation. That capability may support the development of genetic medicines, including therapies that edit regulatory regions of the genome to alter disease risk or treat inherited disorders. By clarifying which genomic variants actually influence gene activity, the tool could make it easier to turn disease-associated genetic findings into actionable therapeutic targets.