MERFISH Mouse Brain Dataset

This dataset provides spatial transcriptomic profiles of the adult mouse brain using MERFISH (Multiplexed Error-Robust Fluorescence In Situ Hybridization), enabling high-resolution cellular gene expression analysis in situ. The MERFISH data consists of 59 coronal sections (10 micron thick) covering the whole anterior to posterior extent of the brain (~200 micron apart) from a single adult male mouse.

It includes both raw and processed versions of the data as streamable Parquet shards. Expression data is aligned to annotated genes, and each cell's expression profile is linked to metadata and spatial information.

Dataset Structure

• raw/: minimally processed gene expression matrix and metadata.
• processed/: normalized and filtered expression matrix.
• gene_metadata.parquet: gene-level annotations.
• cell_metadata.parquet: cell-level annotations.
• expression/: shard-wise Parquet matrices of expression data.

Each sample includes:

• cell_id: unique cell identifier
• expression: float32 vector of gene expression values
• gene_names: ordered list of gene names

Citation

If you use this dataset, please cite:

Xiaoyan Qian, et al.
Spatiotemporal transcriptomic maps of whole mouse brains at single-cell resolution,
Nature (2023).
https://doi.org/10.1038/s41586-023-06808-9

BibTeX:

@article{qian2023spatiotemporal,
  title={Spatiotemporal transcriptomic maps of whole mouse brains at single-cell resolution},
  author={Qian, Xiaoyan and Chang, Yanxiang and Wu, Xinyi and Wang, Youliang and Luo, Chao and others},
  journal={Nature},
  volume={624},
  number={7996},
  pages={343--351},
  year={2023},
  publisher={Nature Publishing Group},
  doi={10.1038/s41586-023-06808-9}
}

License

This dataset is made available under the Creative Commons Attribution 4.0 International (CC BY 4.0) license. You are free to share and adapt the data, provided appropriate credit is given.

How to Load

from datasets import load_dataset

# Load raw version
ds_raw = load_dataset("data4science/merfish", name="raw", streaming=True)

# Load processed version
ds_proc = load_dataset("data4science/merfish", name="processed", streaming=True)