Ceograph

CEll spatial Organization-based graph convolutional network to integrate cell nuclei morphologies and cell-cell spatial interactions in helping clinical dicision, outcome prediction, and model interpretation.

This repository includes

• How to construct graph from cell spatial distribution, how to construct, train, and infer a Ceograph
• How to visualize cellular contribution to Ceograph
• How to visualize morphology and cell-cell interaction contribution to Ceograph

Requirements

• python 3
• torch==1.0.1.post2
• torch-cluster==1.2.4
• torch-geometric==1.0.3
• torch-scatter==1.1.2
• torch-sparse==0.2.5
• torch-vision==0.1.6.dev0
• scikit-learn==0.20.3
• networkx==2.2
• more detailed environment information can be found in Session Info at the bottom of .ipynb files.