ADMETlab: a platform for systematic ADMET evaluation based on a comprehensively collected ADMET database

Table 5 The best classification models for some ADME/T related properties

Property	Method	Features	Fivefold cross validation				External validation dataset
Property	Method	Features	Sensitivity	Specificity	Accuracy	AUC	Sensitivity	Specificity	Accuracy	AUC
HIA	RF	MACCS	0.820	0.743	0.782	0.846	0.801	0.743	0.773	0.831
F (20%)	RF	MACCS	0.731	0.647	0.689	0.759	0.680	0.663	0.671	0.746
F (30%)	RF	ECFP6	0.743	0.605	0.669	0.715	0.751	0.601	0.667	0.718
BBB	SVM	ECFP2	0.962	0.813	0.926	0.948	0.993	0.854	0.962	0.975
Pgp-inhibitor	SVM	ECFP4	0.887	0.789	0.848	0.908	0.863	0.802	0.838	0.913
Pgp-substrate	SVM	ECFP4	0.839	0.807	0.824	0.899	0.826	0.854	0.840	0.905
CYP1A2-inhibitor	SVM	ECFP4	0.833	0.864	0.849	0.928	0.853	0.880	0.867	0.939
CYP1A2-substrate	RF	ECFP4	0.768	0.636	0.702	0.801	0.768	0.637	0.702	0.802
CYP3A4-inhibitor	SVM	ECFP4	0.759	0.858	0.817	0.901	0.788	0.860	0.829	0.909
CYP3A4-substrate	RF	ECFP4	0.798	0.716	0.757	0.835	0.819	0.679	0.749	0.835
CYP2C19-inhibitor	SVM	ECFP2	0.826	0.819	0.822	0.893	0.812	0.825	0.819	0.899
CYP2C19-substrate	RF	ECFP2	0.735	0.744	0.740	0.816	0.871	0.667	0.769	0.853
CYP2C9-inhibitor	SVM	ECFP4	0.719	0.898	0.837	0.900	0.730	0.882	0.830	0.894
CYP2C9-substrate	RF	ECFP4	0.746	0.709	0.728	0.819	0.746	0.709	0.734	0.824
CYP2D6-inhibitor	RF	ECFP4	0.770	0.811	0.793	0.868	0.771	0.812	0.795	0.882
CYP2D6-substrate	RF	ECFP4	0.765	0.73	0.748	0.823	0.792	0.73	0.76	0.833
hERG	RF	2D	0.908	0.700	0.844	0.879	0.888	0.762	0.848	0.873
H-HT	RF	2D	0.780	0.520	0.689	0.710	0.785	0.487	0.681	0.683
Ames	RF	MACCS	0.800	0.841	0.820	0.890	0.848	0.816	0.834	0.897
SkinSen	RF	MACCS	0.685	0.727	0.706	0.760	0.715	0.727	0.731	0.774
DILI	RF	MACCS	0.866	0.813	0.840	0.904	0.830	0.857	0.843	0.910
FDAMDD	RF	ECFP4	0.848	0.812	0.832	0.904	0.853	0.782	0.821	0.892

ISSN: 1758-2946