Table 4.

Performance comparison of all methods on CYP interactions of DS3.

Method	AUC	AUPR	F-measure	Recall	Precision
Substructure-based label propagation model	0.952	0.126	0.206	0.278	0.161
Side-effect-based label propagation model	0.953	0.120	0.199	0.278	0.156
Vilar’s substructure-based model	0.953	0.126	0.196	0.279	0.152
Classifier ensemble method	0.990	0.541	0.553	0.566	0.546
Weighted average ensemble method	0.695	0.484	0.198	0.201	0.201
NDD	0.994	0.830	0.772	0.770	0.775
RF	0.737	0.092	0.161	0.216	0.132
LR	0.977	0.487	0.524	0.589	0.475
Adaptive boosting	0.830	0.143	0.215	0.259	0.185
LDA	0.953	0.327	0.388	0.363	0.425
QDA	0.709	0.317	0.259	0.446	0.184
KNN	0.590	0.064	0.039	0.008	0.190

The best value of each criterion is shown in bold.