Table 2.

Performances of DDI-PULearn and the benchmark methods evaluated by 20 runs of 3-fold cross-validation and 5-fold cross-validation

Evaluation	Method	Precision	Recall	F1-score
3-fold CV	Vilar’s substructure-based method	0.145	0.535	0.229
	Vilar’s interaction-fingerprint-based method	0.377	0.553	0.447
	Zhang’s weighted average ensemble method	0.782	0.703	0.740
	Zhang’s L1 classifier ensemble method	0.788	0.717	0.751
	Zhang’s L2 classifier ensemble method	0.784	0.712	0.746
	DDI-PULearn	0.902	0.822	0.860
5-fold CV	Vilar’s substructure-based method	0.145	0.535	0.229
	Vilar’s interaction-fingerprint-based method	0.377	0.553	0.447
	Zhang’s weighted average ensemble method	0.775	0.659	0.712
	Zhang’s L1 classifier ensemble method	0.785	0.670	0.723
	Zhang’s L2 classifier ensemble method	0.783	0.665	0.719
	DDI-PULearn	0.904	0.824	0.862