Table 9 Ablation analysis on the DrugCombDB dataset
From: DFFNDDS: prediction of synergistic drug combinations with dual feature fusion networks
Method | ACC | BACC | Prec | Rec | F1 | ROC AUC | MCC | Kappa | AP |
|---|---|---|---|---|---|---|---|---|---|
DFFNDDS | 0.871(0.002) | 0.835(0.002) | 0.801(0.008) | 0.747(0.006) | 0.773(0.004) | 0.922(0.003) | 0.684(0.005) | 0.683(0.005) | 0.859(0.006) |
w.o. R-drop | 0.866(0.002) | 0.832(0.006) | 0.785(0.011) | 0.750(0.017) | 0.767(0.007) | 0.919(0.004) | 0.674(0.007) | 0.673(0.008) | 0.854(0.007) |
w.o. fingerprint | 0.841 (0.035) | 0.789 (0.056) | 0.769 (0.050) | 0.663 (0.107) | 0.710 (0.086) | 0.886 (0.054) | 0.606 (0.099) | 0.602 (0.103) | 0.805 (0.086) |
w.o.fine-tuned BERT | 0.859 (0.009) | 0.818 (0.005) | 0.797 (0.016) | 0.713 (0.017) | 0.752 (0.004) | 0.914 (0.002) | 0.657 (0.003) | 0.655 (0.003) | 0.843 (0.004) |
w.o. attention mechanism | 0.860 (0.006) | 0.819 (0.012) | 0.794 (0.014) | 0.719 (0.025) | 0.754 (0.016) | 0.913 (0.008) | 0.659 (0.019) | 0.657 (0.019) | 0.838 (0.014) |
w.o. highway network | 0.857 (0.003) | 0.816 (0.007) | 0.785 (0.008) | 0.714 (0.017) | 0.747 (0.009) | 0.914 (0.003) | 0.649 (0.009) | 0.648 (0.011) | 0.842 (0.008) |
w.o. SMILES | 0.858 (0.005) | 0.815 (0.009) | 0.791 (0.007) | 0.709 (0.019) | 0.748 (0.012) | 0.912 (0.002) | 0.651 (0.013) | 0.649 (0.014) | 0.841 (0.008) |