Table 3 Ablation study on the Davis and KIBA datasets
From: DGDTA: dynamic graph attention network for predicting drug–target binding affinity
Dataset | Methods | GATv2 | GCN | GRU | LSTM | CI | MSE |
|---|---|---|---|---|---|---|---|
Davis | Baseline1 | – | – | – | – | 0.892 | 0.232 |
Model-1 | – | – | – | ✓ | 0.893 | 0.230 | |
Model-2 | ✓ | – | – | – | 0.895 | 0.232 | |
Model-3 | ✓ | – | ✓ | – | 0.896 | 0.228 | |
Model-4 | ✓ | – | – | ✓ | 0.899* | 0.225 | |
Baseline2 | – | ✓ | – | – | 0.881 | 0.245 | |
Model-5 | – | ✓ | – | ✓ | 0.886 | 0.241 | |
Model-6 | ✓ | ✓ | – | – | 0.883 | 0.242 | |
Model-7 | ✓ | ✓ | ✓ | – | 0.887 | 0.239 | |
Model-8 | ✓ | ✓ | – | ✓ | 0.889 | 0.237 | |
KIBA | Baseline1 | – | – | – | – | 0.866 | 0.179 |
Model-1 | – | – | – | ✓ | 0.867 | 0.172 | |
Model-2 | ✓ | – | – | – | 0.869 | 0.175 | |
Model-3 | ✓ | – | ✓ | 0.871 | 0.169 | ||
Model-4 | ✓ | – | – | ✓ | 0.878 | 0.162 | |
Baseline2 | – | ✓ | – | – | 0.891 | 0.139 | |
Model-5 | ✓ | – | ✓ | 0.896 | 0.131 | ||
Model-6 | ✓ | ✓ | – | – | 0.893 | 0.135 | |
Model-7 | ✓ | ✓ | ✓ | – | 0.898 | 0.128 | |
Model-8 | ✓ | ✓ | – | ✓ | 0.902 | 0.125 |