An interactive 3D atlas of gene expression in the mouse brain (v1)

This dataset originates from the re-registration and enhancement of the [Allen Mouse Brain Atlas's coronal in situ hybridization (ISH) data](https://doi.org/10.1038/nature05453). It comprises high-resolution (25 µm isotropic) 3D gene expression volumes for 4,083 genes, generated from 401,660 sections across 1,052 C57BL/6J mouse brains (postnatal day 56). The data were aligned to the [CCFv3bbp reference atlas](http://doi.org/10.1162/imag_a_00565) using an automated pipeline. The pipeline combined [DeepSlice](https://doi.org/10.1038/s41467-023-41645-4) for initial section-to-atlas alignment, [QuickNII](https://doi.org/10.1371/journal.pone.0216796) for manual corrections, and the Advanced Normalization Tools ([ANTs](https://doi.org/10.1101/2024.05.01.592056)) for correction of non-linear distortions. This pipeline achieved accuracy comparable to human novices and significantly outperforming the original ABA registrations. The interpolated volumes reveal fine anatomical details that are not visible at the original 200 µm resolution. From these volumes, a data-driven "CArea" atlas with 55 regions defined by shared molecular signatures was created. This resource is reusable for studies requiring precise spatial gene expression patterns, analysis of molecular brain architecture, and integration with other volumetric datasets in neuroscience.