Table 5 The effect on prediction performance of different graph encoders
Model | Encoder in HG | Encoder in RG | Prediction Performance | ||||
|---|---|---|---|---|---|---|---|
GAT | GCN | GAT | GCN | Jaccard | F1 | PR-AUC | |
\(\text {KAMPNet}_{AC}\) | ✓ | ✗ | ✗ | ✓ | 0.4973 | 0.6454 | 0.7195 |
\(\text {KAMPNet}_{AA}\) | ✓ | ✗ | ✓ | ✗ | 0.4878 | 0.6381 | 0.7160 |
\(\text {KAMPNet}_{CC}\) | ✗ | ✓ | ✗ | ✓ | 0.4869 | 0.6373 | 0.7176 |
\(\text {KAMPNet}_{CA}\) | ✗ | ✓ | ✓ | ✗ | 0.4853 | 0.6358 | 0.7122 |