Copyright (C) 2022 Yuanyuan Zhang, Xiao Wang, Charles Christoffer, & Daisuke Kihara, and Purdue University.

License: GPL v3 for academic use. (For commercial use, please contact us for different licensing.)

Contact: Daisuke Kihara ([email protected])

git clone https://github.com/kiharalab/GNN_pocket
cd GNN_pocket

You have two options to install dependency on your computer, install requirements.txt or install dependencies one by one:

pip install -r requirements.txt 

If you encounter any errors, you can install each library one by one:

click==8.0.3
matplotlib==3.4.3
numpy==1.20.3
pandas==1.3.4
scikit_learn==1.0.2
scipy==1.7.1
torch==1.10.1
tqdm==4.62.3

python3 main.py -h:
  --mode                Running Mode
  --gpu                 Specify the gpu to use
  --test_dir            Specify the directory of original dataset, the structure under this dataset 
                        should be "./id/structure.pqr"
  --test_odir           Specify the directory that you want to save the processed data

python main.py --mode=1 --gpu=1 --test_dir="./dataset/test" --test_odir="./dataset/test_processed"

We strongly recommend you to run on a machine with a SSD, the overall running time for an example would be around 2 minutes with a SSD.

But the running time would be much longer without SSD, it is possible to over 10 minutes in some cases with large structures.

The output files are saved as final_pred which is at the same directory. For example:

cd ./final_pred
cd 23

structure.pqr: the prediction is at column 9, which 1.000 means pocket 1, 0.000 means not a pocket atom.

Two examples about pocktes, the first one is example 23 from test set, the second one is example 142 from validation set.

• [1] Tool: pqr2pdb
• [2] Tool: VisGrid
• [3] Tool: ghecom
• [4] Li, Bin, et al. "Characterization of local geometry of protein surfaces with the visibility criterion." Proteins: Structure, Function, and Bioinformatics 71.2 (2008): 670-683.
• [5] Kawabata, Takeshi. "Detection of multiscale pockets on protein surfaces using mathematical morphology." Proteins: Structure, Function, and Bioinformatics 78.5 (2010): 1195-1211.
• [6] Kipf, Thomas N., and Max Welling. "Semi-supervised classification with graph convolutional networks." arXiv preprint arXiv:1609.02907 (2016).
• [7]Wang, Xiao, Sean T. Flannery, and Daisuke Kihara. "Protein docking model evaluation by graph neural networks." Frontiers in Molecular Biosciences 8 (2021): 402.