GASTON-Mix is a spatial mixture-of-experts (MoE) model for learning domain-specific topographic maps of a tissue slice from spatially resolved transcriptomics (SRT) data.
We will make GASTON-Mix pip
-installable soon. In the meanwhile, you can directly install the conda environment from the environment.yml
file:
conda env create -f environment.yml
Then install GASTON-Mix using pip.
conda activate gaston-mix
pip install -e .
Installation should take <10 minutes.
See our tutorial tutorial.ipynb
and check out our readthedocs.
- torch
- matplotlib
- pandas
- scikit-learn
- numpy
- jupyterlab
- seaborn
- tqdm
- scipy
- scanpy
See full list in environment.yml
file. GASTON-Mix can be run on CPU or GPU.
We note that GASTON-Mix sometimes uses clusters from CellCharter to initialize its gating network. We suggest either making a separate environment to run CellCharter and follow their tutorial, or using a different initialization (see tutorial).
The GASTON-Mix pre-print is available on biorXiv. If you use GASTON-Mix for your work, please cite our paper.
@article {Chitra2025,
author = {Chitra, Uthsav and Dan, Shu and Krienen, Fenna and Raphael, Benjamin J.},
title = {GASTON-Mix: a unified model of spatial gradients and domains using spatial mixture-of-experts},
elocation-id = {2025.01.31.635955},
year = {2025},
doi = {10.1101/2025.01.31.635955},
publisher = {Cold Spring Harbor Laboratory},
eprint = {https://www.biorxiv.org/content/early/2025/02/04/2025.01.31.635955.full.pdf},
journal = {bioRxiv}
}
For questions or comments, please file a Github issue and/or email Uthsav Chitra (uchitra@broadinstitute.org)