SCALE: Unsupervised Multi-Scale Domain Identification in Spatial Omics Data

SCALE is a Python package for identifying multi-scale spatial domains in spatial omics data. It leverages graph neural representation learning and an entropy-based search algorithm to detect stable spatial domains at different scales, enabling comprehensive analysis of tissue organization. The preprint can be found here.

Installation

Clone the repository:

git clone https://github.com/imsb-uke/scale.git
cd scale

We recommend to use poetry to install the package.

poetry install

Otherwise you can install the package via pip:

pip install -e .

Quick Start

You can find a short vignette here notebooks/vignette.ipynb, applying SCALE to MERFISH mouse brain data. For convenience the data is provided in the data folder.

Citation

If you use SCALE in your research, please cite:

@article{yousefi2025scale,
  title={SCALE: Unsupervised Multi-Scale Domain Identification in Spatial Omics Data},
  author={Yousefi, Behnam and Schaub, Darius P and Khatri, Robin and Kaiser, Nico and Kuehl, Malte and Ly, Cedric and Puelles, Victor G and Huber, Tobias B and Prinz, Immo and Krebs, Christian F and others},
  journal={bioRxiv},
  pages={2025--05},
  year={2025},
  publisher={Cold Spring Harbor Laboratory}
}