The repository contains the pytorch implementation of SigGraInferNet.
python 3.6
pytorch 1.6.0 or above
scikit-learn
tensorboard
tqdm
we also provide conda virtual environment for you. You can create a new terminal and cd to the root directory of repository and run following code:
conda env create -f environment.yml
A new virtual environment called SigGraInferNet will be created. You can activate the environment by:
conda activate SigGraInferNet
Unfortunately we cannot provide the dataset in the repository, you can download corresponding dataset here:
STRING: https://string-db.org/
Hallmark gene sets:https://www.gsea-msigdb.org/gsea/msigdb/collections.jsp
TCGA dataset:https://xenabrowser.net/datapages/
The KEGG pathway is get and processed by R with package graphite
After you have all the data, we dicuss the detail of how to process the data in the paper.
First, modify the data path in args.py
and corresponding pytorch data processing function in util.py
based on your data format. Then run the following code to start training:
python train.py -n=SigGraInferNet
This will run the SigGraInferNet with default parameter setting. You can also change the parameters in the args.py
. You can also visualize the training process in tensorboard:
tensorboard --logdir=save --port=5678