Single-cell machine learning uncovers genetically anchored, cell-type specific programs of Alzheimer's disease
Journal:
medRxiv
Published Date:
Feb 6, 2026
Abstract
Aging and genetic risk shape the molecular programs that confer cellular vulnerability in Alzheimer's disease (AD), but whether these programs differ between clinical symptoms and neuropathological burden remains unclear. Using single nucleus RNA sequencing (snRNA seq) from the dorsolateral prefrontal cortex, we applied TriSCOPE, integrating multivariate predictive modeling, differential expression, and pseudotime inference, across 2,797,869 nuclei from 544 donors spanning four human AD cohorts. Across six major brain cell types and 41 fine grained subclusters, we identified robust pan cell type and cell type specific programs, including ARL17B and glial stress/immune candidates such as CRYAB (oligodendrocytes) and IFI44L (microglia), alongside previously reported neuronal risk-associated genes. Genes prioritized by predictive modeling were then evaluated using conditional GWAS eQTL colocalization, providing additional evidence that inherited Alzheimer's disease risk and cell type specific transcriptional regulation may converge at a subset of these loci. Pseudotime analysis showed that a select subset of genes identified through predictive modeling and differential expression also ranked among the most influential drivers of disease associated transcriptional trajectories, providing independent trajectory level support for these candidates. By jointly modeling genetic, demographic, and transcriptomic axes, our study prioritizes genetically supported, cell type resolved molecular candidates associated with Alzheimer's disease, providing a framework for mechanistic follow-up and hypothesis-driven therapeutic investigation in age-related neurodegeneration.
