Combinatorial prediction of therapeutic perturbations using causally-inspired neural networks

Guadalupe Gonzalez*, Xiang Lin*, Isuru Herath, Kirill Veselkov, Michael Bronstein, and Marinka Zitnik

Project structure

The project consists of next folders:

• data contains all of the data on which our models were built. On how to obtain this data, refer to Data section,
• docs contains documentation, built with 'sphinx',
• src/pdgrapher contains the source code for PDGrapher,
• examples contains some examples that demonstrate the use of PDGrapher library, read the README.md in examples folder for more instructions,
• tests contains unit and integration tests.

Virtual environment

conda env create -f conda-env.yml
conda activate pdgrapher
pip install pip==23.2.1
pip install -r requirements.txt

pip install torch==1.10.1+cu111 torchvision==0.11.2+cu111 torchaudio==0.10.1 -f https://download.pytorch.org/whl/cu111/torch_stable.html
pip install torch-scatter==2.0.9 -f https://data.pyg.org/whl/torch-1.10.1+cu111.html
pip install torch-sparse==0.6.12 -f https://data.pyg.org/whl/torch-1.10.1+cu111.html
pip install torch-cluster==1.5.9 -f https://data.pyg.org/whl/torch-1.10.1+cu111.html
pip install torch-spline-conv==1.2.1 -f https://data.pyg.org/whl/torch-1.10.1+cu111.html
pip install torch-geometric==2.0.4 -f https://data.pyg.org/whl/torch-1.10.1+cu111.html
pip install torchmetrics==0.9.3
pip install lightning==1.9.5

Data

For processed data, download the compressed folders and place them in data/processed/ with the following commands:

cd data
mkdir processed
cd processed
wget https://figshare.com/ndownloader/files/43624557
tar -xzvf torch_data.tar.gz
cd ../
wget https://figshare.com/ndownloader/files/43632327
tar -xzvf splits.tar.gz

Building

This project can be build as a Python library by running pip install -e . in the root of this repository.

Documentation

Documentation can be build with next two commands:

• sphinx-apidoc -fe -o docs/source/ src/pdgrapher/ updates the source files from which the documentation is built,
• docs/make html builds the documentation

Then, the documentation can be accessed locally by going to docs/build/html/index.html All of the settings along with links to instructions can be found and modified in docs/source/conf.py.

Notebooks

Train PDGrapher on chemical dataset jupyter notebook
Train PDGrapher on genetic dataset jupyter notebook
Test PDGrapher on chemical/genetic dataset jupyter notebook

Reference

@article{gonzalez2025combinatorial,
title={Combinatorial prediction of therapeutic perturbations using causally-inspired neural networks},
author={Gonzalez, Guadalupe and Lin, Xiang and Herath, Isuru and Veselkov, Kirill and Bronstein, Michael and Zitnik, Marinka},
journal={bioRxiv},
pages={2024--01},
year={2025}
}

License

The code in this package is licensed under the MIT License.

Questions

Please leave a Github issue or contact Guadalupe Gonzalez, Xiang Lin, or Marinka Zitnik