Researchers at the Allen Institute and UCSF developed CellTransformer, an AI neural network that classified 3.9 million mouse brain cells across >5,000 neuron types to produce a data‑driven brain atlas with roughly 1,300 regions, published in Nature Communications. The model uses spatial transcriptomics and single‑cell gene signatures to define fine anatomical subdivisions beyond prior atlases, enabling more precise localization of cell types linked to behavior and disease. Authors and external experts suggested the tool could localize where disease‑causing mutations exert effects, improving mechanistic studies and target validation. The result represents a significant advance in computational neuroanatomy and spatial biology.