Reproducibility repository for Multi-GASTON applied on spatial metabolomics data.
This repository provides two jupyter notebook tutorials on applying the deep learning model Multi-GASTON—-referred to here as Metabolic Topography Mapper: MET-MAP-—to spatial metabolomics data from murine liver and small intestine. To support reproducibility, we include example datasets, neural network outputs, and downstream metabolite analyses. Although these tutorials focus on metabolomics, MET-MAP is broadly applicable to any spatially-resolved data, enabling the recovery of tissue architecture and spatial patterns of feature variation across diverse organs. For more details about the model and analysis, please visit: https://github.com/raphael-group/Multi-GASTON and paper pre-print https://www.biorxiv.org/content/10.1101/2025.06.02.657306v1.
For enviroment setup, please refer to Multi-GASTON installation at https://github.com/raphael-group/Multi-GASTON/tree/main. After installing Multi-GASTON package, simply activate the conda enviroment required for the jupyter notebooks.
conda activate multi_gaston_env
For liver and small intestine metabolomics data used in the paper, please refer to Figshare repositories:
Liver spatial metabolomics: https://doi.org/10.6084/m9.figshare.29318279.v1
Intestine spatial metabolomics: https://doi.org/10.6084/m9.figshare.29318342.v1