Shaow/mousebrain_merfish_zhuang

    # Whole Mouse Brain MERFISH · Zhuang ABCA-4

    Curated, ready-to-load spatial transcriptomics dataset.

    ## Source

    - Paper: [Zhang, Jung, Zhuang et al., Nature 2023](https://www.nature.com/articles/s41586-023-06808-9)
    - Canonical download: Brain Image Library doi 10.35077/act-bag · counts_mouse4_sagittal subset

    ## Scale

    | Property | Value |
    |---|---|
    | Technology | MERFISH (1147-gene panel) |
    | Species | Mus musculus |
    | Tissue | Whole mouse brain (sagittal sections) |
    | Sections / slices | 3 |
    | Total cells / spots | 173,849 |

    ## Files

    - `counts_mouse4_sagittal.h5ad`

• counts_mouse4_sagittal_normalized.h5ad

    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)
  
    `sample_id`, `slice_id`, `fov`, `volume`, `center_x`, `center_y`
  
    ## Notes
  
    Zhuang-ABCA-4 mouse-brain MERFISH atlas, sagittal sections of mouse 4 (173,849 cells × 1,147 genes total, 3 sections at 43.8K / 52.3K / 77.7K cells each). `counts_mouse4_sagittal.h5ad` is the raw integer-count version (X max=151). `counts_mouse4_sagittal_normalized.h5ad` keeps the originally-distributed normalized version as backup. `obsm['spatial']` is set from `obs[center_x, center_y]`.
  
    ## Usage
  
    ```python
  

import scanpy as sc from huggingface_hub import snapshot_download d = snapshot_download(repo_id='Shaow/mousebrain_merfish_zhuang', repo_type='dataset', allow_patterns='counts_mouse4_sagittal.h5ad') adata = sc.read_h5ad(f'{d}/counts_mouse4_sagittal.h5ad')

filter to one section if desired

adata_one = adata[adata.obs['slice_id'] == 'sa2_slice2'].copy() # ≈ 52,298 cells

        ## 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.