Latent Factor Modeling Reveals Unexpected Spatial Heterogeneity in Human Alzheimer's Disease Brain Transcriptomes

Alzheimer's disease is characterized by complex molecular and cellular heterogeneity, which complicates efforts to identify consistent biomarkers and therapeutic targets. To better characterize the heterogeneity, we applied latent factor modeling to RNA sequencing data from approximately 2,500 human Alzheimer's disease brain samples, uncovering underlying patterns in gene expression. These transcriptional groups demonstrated unique gene expression profiles related to synaptic and neuronal pathways, vasculature development, and protein folding and antigen processing. Notably, this latent factor reflects variation in spatial sampling. Adjusting for the latent factor improved the identification of differentially expressed genes in disease samples. This finding suggests that spatial heterogeneity is a pervasive driver of transcriptomic variation and has important implications for future studies of Alzheimer's disease and related neurological disorders.