Researchers at the University of California, Irvine report that a machine-learning system called SIGNET can infer cause-and-effect links between genes in human brain tissue, revealing extensive rewiring of gene regulation—especially in excitatory neurons—in Alzheimer’s disease.
A team led by Min Zhang and Dabao Zhang at the University of California, Irvine’s Joe C. Wen School of Population & Public Health has produced what it describes as highly detailed maps of how genes influence one another in brain cells affected by Alzheimer’s disease, using a machine-learning platform called SIGNET.
To build the maps, the researchers analyzed molecular data from donated human brain tissue from 272 participants enrolled in two long-running aging studies: the Religious Orders Study and the Rush Memory and Aging Project (often jointly referred to as ROSMAP). The approach integrates single-cell (single-nucleus) RNA sequencing with matched subject-level genetic variation data, allowing the team to move beyond gene-to-gene correlations and infer likely directional, causal regulatory relationships.
Using SIGNET, the researchers constructed causal gene regulatory networks for six major brain cell types. The largest set of inferred regulatory relationships appeared in excitatory neurons. In the underlying study materials, the excitatory-neuron network contained 5,910 inferred “regulations,” a scale the authors say points to extensive rewiring of gene regulation as Alzheimer’s progresses.
Min Zhang, a co-corresponding author and professor of epidemiology and biostatistics, said that while different brain cell types are known to play distinct roles in Alzheimer’s, the molecular-level relationships have been difficult to untangle. She said the new work provides cell-type-specific maps intended to shift the field from observing correlations to identifying mechanisms that may actively drive disease progression.
The work also highlighted “hub genes” that appear to act as central regulators within the networks. The researchers reported previously underappreciated regulatory roles for well-known Alzheimer’s-related genes such as APP, including effects in inhibitory neurons.
To bolster confidence in the findings, the team reported validating key patterns using an independent set of human brain samples. The researchers said the same framework could be applied to other complex diseases, including cancer, autoimmune disorders and mental health conditions.
The findings were reported by UC Irvine and published in Alzheimer’s & Dementia: The Journal of the Alzheimer’s Association, according to the university’s summary of the study, which lists the paper as appearing in 2026 (volume 22, issue 2) with DOI: 10.1002/alz.71053. The ScienceDaily summary attributes partial funding support to the National Institute on Aging and the National Cancer Institute.
Alzheimer’s disease is the leading cause of dementia and is projected to affect nearly 14 million Americans by 2060, according to the UC Irvine summary.
