DGMP: Identifying Cancer Driver Genes by Jointing DGCN and MLP from Multi-omics Genomic Data

by Shaowu Zhang* (zhangsw@nwpu.edu.cn), Jingyu Xu, Tong Zhang

Requirements

Our project is developed using PyTorch 1.7.1 in Python 3.7. We recommend you to use anaconda for dependency configuration.

First create an anaconda environment called DGMP by

conda create -n DGMP python=3.7
conda activate DGMP

Then, you need to install torch manually to fit in with your server environment (e.g. CUDA version 11.0). Run:

conda install pytorch==1.7.1 torchvision==0.8.2 cudatoolkit=11.0

Besides, torch-scatter and torch-sparse are required for dealing with sparse graph, which can be installed using the following commands:

pip install torch-sparse==0.6.8 -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
pip install torch-scatter==2.0.6 -f https://pytorch-geometric.com/whl/torch-1.7.1+cu110.html
pip install torch-geometric==1.6.3

Run

Training and testing our DGMP model with the following commands:

cd code
python DGMP.py --gpu-no 0 --dataset DawnNet

Note: In CPU computing scenarios, please assignment '-1' to '--gpu-no' parameter

Result

The output of our model is supplied at './code/result/', which consist with two variates 'logits' and 'acc'. The probability of each gene is non-cancer driver gene (the first column) or cancer driver gene (the second column) have been stored in 'logits', and the mean accuracy of 5-CV are in 'acc'

License

DGMP is released under the MIT License. See the LICENSE file for more details.

Acknowledgements

The template is borrowed from Pytorch-Geometric benchmark suite. We thank the authors of following works for opening source their excellent codes. DiGCN