TENET

TENET: Triple-Enhancement based Graph Neural Network for Cell-cell Interaction Network Reconstruction from Spatial Transcriptomics

TENET Installation

Create the conda environment (Default installation path), other installation path, use -p to select your own path:

conda env create -f environment.yml

List all of the environment:

conda info -envs

To activate the environment:

conda activate TENET

If the environment cannot be installed successfully, follow the following instructions:

conda create --name TENET python=3.8

Activate the Python environment, and start installing the packages.

conda activate TENET
pip install -r requirements.txt

Installing pytorch torchvision torchaudio separately (check your own cuda version), my Cuda version is 11.7.

# To check the Cuda version
nvcc -V

If the Cuda version is different from mine, the following url and the provided .whl files should be different.

pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117
# When installing pytorch dependencies, they are difficult to install and have a long wait time, my recommendation is to download their.whl file and install it.
pip install xxx.whl

Training

To get started, run the following code (all the model python files are placed at the "TENET" directory, not the current directory):

cd TENET
python main.py

-m (mode; default to be "train")
-t (train-test-split ratio)
-fp, -fn (two noise ratio)

another hyperparamters can be modify by yourself.

Cite

@article{lee2024tenet,
  title={TENET: Triple-Enhancement based Graph Neural Network for Cell-cell Interaction Network Reconstruction from Spatial Transcriptomics},
  author={Lee, Yujian and Xu, Yongqi and Gao, Peng and Chen, Jiaxing},
  journal={Journal of Molecular Biology},
  pages={168543},
  year={2024},
  publisher={Elsevier}
}