The Allen Gene Expression ISH dataset has been a highly impactful resource for Neuroscientists. Mapping the location within the brain of more than 20,000 genes, it has transformed our understanding of the relationship between gene expression and neuroanatomical structures. Nevertheless, at nearly 20 years old, the accuracy of aligning or matching the ISH section images to the corresponding regions or structures in the atlas was limited by the registration technology available at the time. As a result, the registrations provided by the Allen Institute are accurate to approximately 300 μm. Here we re-register the coronal subset of this dataset to achieve a considerably higher registration accuracy. Using a deep learning registration method, DeepSlice (Carey et al., Nature Communications, 2023, PMID: 37735467), we have created registrations with 10 μm accuracy. Furthermore, we have applied a non-linear registration step to correct for tissue deformations. These registrations will be made publicly available with viewer links via the EBRAINS Knowledge Graph data repository. In the future we aim to interpolate these section datasets, creating high resolution maps of whole brain gene expression in three-dimensions. Funded from the European Union’s Research and Innovation Program Horizon Grant Agreement No. 101147319 (EBRAINS 2.0).