Shaow/mousebrain_visium_3section

# Mouse Brain Visium · 3 complementary sections

Curated, ready-to-load spatial transcriptomics dataset.

## Source

• —Paper: 10x Genomics public datasets
• —Canonical download: cf.10xgenomics.com/samples/spatial-exp/1.1.0/V1_*

## Scale

## Files

• —Visium_sagittal-anterior2/V1_Mouse_Brain_Sagittal_Anterior_Section_2_filtered_feature_bc_matrix.h5
• —Visium_sagittal-posterior2/V1_Mouse_Brain_Sagittal_Posterior_Section_2_filtered_feature_bc_matrix.h5
• —Visium_coronal/V1_Adult_Mouse_Brain_filtered_feature_bc_matrix.h5

Each .h5ad follows the AnnData spec:

• —.X — gene expression matrix (cells × genes), sparse where natural
• —.obs — per-cell annotations (see "Metadata" below)
• —.obsm['spatial'] — (n_cells, 2) float32 spatial coordinates
• —(where present) .layers['count'] — raw integer counts
• —(where present) .obsm['spatial3d'] — (n_cells, 3) float32 (x, y, z=section)

## Metadata (obs columns)

in_tissue, array_row, array_col

## Notes

Three 10x Visium mouse-brain serial sections offering complementary tissue views. Original 10x outputs (filtered h5 + spatial/) kept in conventional subfolders so sc.read_visium(folder) works directly. After filtering to in_tissue==1: 2825 / 3289 / 2702 spots respectively.

## Usage

import scanpy as sc
from huggingface_hub import snapshot_download
d = snapshot_download(repo_id='Shaow/mousebrain_visium_3section', repo_type='dataset')
adata_st_list = []
for sub, h5 in [
    ('Visium_sagittal-anterior2',  'V1_Mouse_Brain_Sagittal_Anterior_Section_2_filtered_feature_bc_matrix.h5'),
    ('Visium_sagittal-posterior2', 'V1_Mouse_Brain_Sagittal_Posterior_Section_2_filtered_feature_bc_matrix.h5'),
    ('Visium_coronal',             'V1_Adult_Mouse_Brain_filtered_feature_bc_matrix.h5'),
]:
    a = sc.read_visium(f'{d}/{sub}', count_file=h5)
    a.var_names_make_unique()
    adata_st_list.append(a[a.obs['in_tissue'] == 1].copy())

## Citation

If you use this dataset, please cite the source paper above.

## License

MIT for the curation/preparation. Underlying data inherits the license of the upstream publication (typically CC-BY-4.0); please see the source paper.