Shaow/mouseembryo_seqfish_stereoseq_pair

    # Mouse Embryo · paired seqFISH + Stereo-seq slice

    Curated, ready-to-load spatial transcriptomics dataset.

    ## Source

    - Paper: [Lohoff et al., Nat. Biotechnol. 2022 (seqFISH); Chen et al., Cell 2022 MOSTA (Stereo-seq)](https://www.nature.com/articles/s41587-021-01006-2)
    - Canonical download: seqFISH: SpatialMouseAtlas; Stereo-seq: MOSTA STDS0000058 (E9.5_E1S1)

    ## Scale

    | Property | Value |
    |---|---|
    | Technology | seqFISH (351 genes) + Stereo-seq (full transcriptome) |
    | Species | Mus musculus |
    | Tissue | Mouse embryo (E8.5 + E9.5) |
    | Sections / slices | 2 |
    | Total cells / spots | 20,098 |

    ## Files

    - `h5ad/adata_seqfish.h5ad`

• h5ad/adata_stereoseq.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)
  
    `embryo (seqFISH)`, `celltype_mapped_refined (seqFISH)`, `annotation (Stereo-seq)`
  
    ## Notes
  
    Cross-platform pair from two studies: seqFISH embryo2 (14,185 cells × 351 genes) + MOSTA E9.5_E1S1 Stereo-seq (5,913 spots × 25,568 genes). 347 genes are shared between the platforms. Use as a benchmark for cross-technology spatial integration.
  
    ## Usage
  
    ```python
  

import scanpy as sc from huggingface_hub import snapshot_download d = snapshot_download(repo_id='Shaow/mouseembryo_seqfish_stereoseq_pair', repo_type='dataset') adata_seqfish = sc.read_h5ad(f'{d}/h5ad/adata_seqfish.h5ad') adata_stereoseq = sc.read_h5ad(f'{d}/h5ad/adata_stereoseq.h5ad')

        ## 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.