spMetaTME

spMetaTME: Leveraging self-supervised graph neural networks to decode spatial metabolic communities and metabolic interactions for unveiling tumour heterogeneity within the tumour microenvironment

• PyPI package: https://pypi.org/project/spmetatme/
• Free software: MIT License
• Documentation: https://spmetatme.readthedocs.io.

Create environment

pip install torch==2.7.1 torchvision==0.22.1 torchaudio==2.7.1 --index-url https://download.pytorch.org/whl/cu126
pip install pyg_lib torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.7.0+cu126.html
pip install torch_geometric

Unified CLI Usage

The spmetatme package provides a unified command line interface for your analysis pipeline.

• Detailed Instructions: CLI Guide
• Manual: spmetatme --help

1. Infer Fluxes (Basic Example)

spmetatme infer --input data/BC_515_Section_1.h5ad --model-species human --metabolic-model breast_cancer

2. Pretrain (Basic Example)

spmetatme pretrain --training-list data/*.h5ad --metabolic-model breast_cancer --repo-id MyRepo/v1

Docker Usage

We provide a full-featured Docker image that includes all dependencies (including R's mclust for domain identification).

• Detailed Instructions: Docker Guide

Using the Docker Helper (Windows)

# Run inference
run_docker.cmd infer --input data/BC_515_Section_1.h5ad --model-species human --metabolic-model breast_cancer

# Run pretraining
run_docker.cmd pretrain --training-list data/*.h5ad --metabolic-model breast_cancer --repo-id MyRepo/v1

Manual Docker Command (Linux / Mac)

docker run --rm -it --gpus all -v $(pwd):/app spmetatme infer --input data/BC_515_Section_1.h5ad --model-species human --metabolic-model breast_cancer

Please cite

Verma, S., Doan, L. M. T., Kang, J. H., Colombo, G., Serganov, I., Zappasodi, R., Occhipinti, A., & Angione, C. spMetaTME: A spatial atlas of tumour microenvironment metabolism and interactions.