Prediction of Phase-Separating Protein Classifier by GCN
To run this program, you may need:
- Python 3.6 or later
- Pytorch 1.12.1 and other related packages
- Windows 10 enviroment
- GPU (optional for cuda)
- Set up your enviroment and download the code from github:
git clone https://github.com/ken0414/Graph-PS.git
- Put your data into the appropriate folder:
[protein structure feature] --> ./data/graph
[protein node feature(PSSM)] --> ./data/pssm
- Activate your enviroment and run main.py:
$ python main --mode cv --run 10
Within this line of code, you can choose between cross-validation mode or independent test set mode by modifying the value after mode, n2v to change the file used, and run to select the number of repetitive runs.
option | value |
---|---|
mode |
cv orout |
run |
int value for run times |
- Get result:
For each fold in a single cv, you can get the best epoch of the train in
train_result.txt
. After all fold trained in a single cv, you can get the evaluation of all fold inpredict_result.txt
and the result of prediction in fold./result
. If you run onout
mode, there will be only 1 result intrain_result.txt
andpredict_result.txt
. If the value ofrun
is bigger than 1, former result intrain_result.txt
of a singlecv
orout
will be override. If you want to save the result of this file, please modify the code on your own. Samely, if you run the main.py again, all the result will be override.