Shaow/mouseembryo_stereoseq_3d

    # Mouse Embryo Stereo-seq · 5 adjacent E16.5 sections

    Curated, ready-to-load spatial transcriptomics dataset.

    ## Source

    - Paper: [Chen et al., Cell 2022 (MOSTA)](https://www.cell.com/cell/fulltext/S0092-8674(22)00399-3)
    - Canonical download: https://db.cngb.org/stomics/mosta · STDS0000058 · embryo 2 sections 8–12

    ## Scale

    | Property | Value |
    |---|---|
    | Technology | Stereo-seq (DNA-nanoball arrays) |
    | Species | Mus musculus |
    | Tissue | Mouse embryo at E16.5 |
    | Sections / slices | 5 |
    | Total cells / spots | 568,437 |

    ## Files

    - `h5ad/E16.5_E2S8.MOSTA.h5ad`

• h5ad/E16.5_E2S9.MOSTA.h5ad

• h5ad/E16.5_E2S10.MOSTA.h5ad

• h5ad/E16.5_E2S11.MOSTA.h5ad

• h5ad/E16.5_E2S12.MOSTA.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)
  
    `annotation`
  
    ## Notes
  
    5 adjacent E16.5 sagittal sections from MOSTA embryo 2 (E2S8..E2S12). Total 568,437 cells. Suitable for 3D tissue-architecture reconstruction. Same X / layers convention as the temporal release: `.X` holds log-normalized values, `.layers['count']` holds raw int counts.
  
    ## Usage
  
    ```python
  

import scanpy as sc from huggingface_hub import snapshot_download d = snapshot_download(repo_id='Shaow/mouseembryo_stereoseq_3d', repo_type='dataset') slices = ['E16.5_E2S8', 'E16.5_E2S9', 'E16.5_E2S10', 'E16.5_E2S11', 'E16.5_E2S12'] adata_st_list = [] for s in slices: a = sc.read_h5ad(f'{d}/h5ad/{s}.MOSTA.h5ad') a.X = a.layers['count'] adata_st_list.append(a)

        ## 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.