Multiple PPI Transformer (MPIT) is based on Graph Transformer Networks (GTNs) to discern specific cancer driver genes. MPIT adeptly amalgamates data from diverse PPI and multi-omics sources through the synergy of embedding alignment and fusion.
We recommend using conda to configure the code runtime environment:
conda create -n mpit python=3.8.12
conda install pytorch==1.12.1 -c pytorch
pip install torch_geometric==2.3.0 transformers wandb
Pytorch versions that allow torch_geometric >= 2.3.0 are okay.
We recommend getting MPIT using Git from our Github repository through the following command:
git clone https://github.com/NBStarry/Multiple-PPI-Transformer.git
To verify a successful installation, just run:
unzip data/Lung_Cancer_Matrix/A549.zip
python main.py -g 0 -t -d # You need to select an idle GPU using "-g".
| Item | Details |
|---|---|
| System | Ubuntu 20.04.1 LTS |
| RAM Memory | 256G |
| GPU Memory | NVIDIA GeForce RTX, 24G |
| Train Time | ~ 5h |
The above table reports our computing details during MPIT development and IS NOT our computing requirement.
If your computer does not satisfy the above, you may try to lower down the memory used during model training by reduce the sampling parameters, the batch size or so on.
If you are having problems with our work, please use the Github issue page.