a method for CPI and DTA prediction.
The first thing to do for WGNN_DTA is to construct the protein weighted graph, we prepare a script to do this, just run it:
python graph_prepare.py
And the script can generate graphs for 4 datasets, which are human, celegants, davis and kiba, just simply edit the dataset name in the script.
python train_cpi.py 0 0 0
where there are 3 parameters: first parameter: datasets, and 0 for human, 1 for celegants; second parameter: gpu number, the selected gpu, change the script if you have more gpus than two; third parameter: ratio, the ratio of negative and positive, 0 for 1:1, 1 for 3:1 and 2 for 5:1.
python train_dta.py 0 0
where there are 2 parameters: first parameter: datasets, and 0 for davis, 1 for kiba; second parameter: gpu number, the selected gpu, change the script if you have more gpus than two.
There several models trained by us in models directory, just set the epoch = 0 to reproduce our result.
The codes are based on GNN, which are easy to entend to your demands (even struture-based methods). You can change the codes according to your needs.