Youzhe

We constructed a harmonized neurovascular transcriptomic resource from the NVU_hyz dataset collection. The processed resource contained 138 sample- or section-level entries projected onto a unified 60,607-gene vocabulary and normalized using library-size scaling followed by log1p transformation. The single-cell/nucleus RNA-seq component comprised 14 dataset-level entries and 11,496,206 cells/nuclei, including public accession-level brain single-cell resources, an AD hippocampal single-nucleus cohort, and a cortical single-nucleus cohort. The spatial transcriptomic component comprised 124 tissue sections or chips and 5,673,194 spatially resolved cells/spots across three datasets: AD cortical spatial transcriptomics (12 sections, 526,904 cells/spots), AD hippocampal spatial transcriptomics (32 chips, 1,867,759 cells/spots), and cortical spatial transcriptomics (80 chips, 3,278,531 cells/spots). This harmonized data structure enabled cross-dataset representation learning, spatial transfer, and sample-level AD/Control readout analyses in a shared gene space.