Table 4.

AUPR for 10 × 5-fold cross-validation over Data-1 (P-values of paired t-test against DrugE-Rank with all features)

Method	Drug-based interactions	Target-based interactions
k-NN	$0.3490 (1.97\times {10}^{-34})$	$0.4569 (2.24\times {10}^{-14})$
BLM-svc	$0.3712 (7.69\times {10}^{-32})$	$0.5138 (6.23\times {10}^{-09})$
BLM-svr	$0.3864 (7.25\times {10}^{-30})$	$0.5631 (1.30\times {10}^{-03})$
LapRLS	$0.3848 (1.01\times {10}^{-33})$	$0.5517 (9.55\times {10}^{-04})$
NetLapRLS	$0.3769 (2.10\times {10}^{-34})$	$0.5124 (8.84\times {10}^{-10})$
WNN-GIP	$0.3641 (1.38\times {10}^{-29})$	$0.5281 (2.03\times {10}^{-05})$
RF	$0.3418 (8.35\times {10}^{-29})$	$0.5014 (1.80\times {10}^{-10})$
GBDT	$0.3328 (1.07\times {10}^{-29})$	$0.5278 (1.09\times {10}^{-06})$
DrugE-Rank	$0.3874 (1.44\times {10}^{-21})$	$0.5427 (7.57\times {10}^{-05})$
(DFV and TFV)
DrugE-Rank	$0.4853 (0.065)$	$0.5841 (0.51)$
(PFV only)
DrugE-Rank	$\mathbf{0}.\mathbf{4917}$	$\mathbf{0}.\mathbf{5906}$
(all features)

The AUPRs achieved by the best model, DrugE-Rank (all features), are highlighted in bold face.