From: End-to-end learning for compound activity prediction based on binding pocket information
Hyperparameter | Value/method |
---|---|
Dimensions of feature vector | 10 |
Compound r-radius subgraphs | 2 |
Layers of compound graph neural network | 3 |
Protein pocket r-radius subgraphs | 1 |
Layers of protein pocket graph neural network | 3 |
Learning rate | 0.001 |
Learning rate decay | 0.5 |
Decay interval | 10 |
Optimization function | Adam |
Epoch size | 100 |
Batch size | 1 |