Official implementation of MultiHyperGNN
Running environment: Python 3.9 Pytorch 1.8.1 networkx 2.5 dgl 0.9.1 pandas 1.4.2 numpy, scipy
Dataset:
- Genes:
a. Input modes:
Expression: expr_in_muscle.csv, expr_in_whole_blood.csv
Graph: graph_in_muscle.csv, graph_in_whole_blood.csv
b. Output modes:
Expression: expr_out_lung.csv, expr_out_skin_not_sun_exposed.csv, expr_out_skin_sun_exposed.csv
Graph: graph_out_lung.csv, graph_out_skin_not_sun_exposed.csv, graph_out_skin_sun_exposed.csv
- Temperature:
a. Early morning:
Value: t1.csv
Graph: c1.csv
b. Late morning:
Value: t2.csv
Graph: c2.csv
c. Afternoon:
Value: t3.csv
Graph: c3.csv
d. Night:
Value: t4.csv
Graph: c4.csv
Train the model:
- To train the model, run: python trian.py.
a. Note that train.py is written specifically for Genes, to train on Temperature, please replace with the corresponding Temperature data in both train.py and test.py.
b. Trained model will be automatically saved as "model.pt".
c. To change model parameters, please directly modify the parameters of model() in both train.py and test.py.
d. To change training epochs, please change the value of "epochs" in train.py.
- To test the model, run: python test.py.