A human brain single-cell and single-nucleus transcriptomic resource across age intervals

It was constructed from experimentally generated scRNA-seq/snRNA-seq data derived from non-disease human brain samples and does not include disease samples, disease-model samples, organoid-derived datasets, simulated datasets, or statistically inferred single-cell expression profiles. The resource harmonizes 22 published human brain scRNA-seq/snRNA-seq datasets and spans 15 age intervals from embryonic and fetal development to childhood, adolescence, adulthood, and late adulthood.

The dataset contains 2,602,031 cells or nuclei, 24,659 genes, 286 donors, 448 biological samples, 3,744 library- or source-level records, and 48 brain regions. All records were processed through unified metadata curation, age-interval mapping, brain-region harmonization and sex-field standardization, and four integration representations were built (raw PCA, Harmony, scVI and RPCA), resulting in 75 clusters and 12 major cell types.The released package includes 10X-compatible expression matrices split by source dataset, cell-level metadata (21 fields), 50-dimensional coordinates and matching 2D UMAPs for the four representations, per-cell LISI validation, source-dataset provenance and licence information, file manifests and MD5 checksums, and example R and Python reader scripts (bundled in scripts/readers.zip).