FP-ADMET: a compendium of fingerprint-based ADMET prediction models

Table 5 Performance metrics for the best performing fingerprint-based regression models

Endpoint	FP	Calibration			Validation
Endpoint	FP	R²	RMSE	MAE	R²	RMSE	MAE
\(\log \ S\)	PUBCHEM	0.77	1.15	0.81	0.78	1.12	0.78
Intrinsic clearance (\(CL_{int}\))	RAD2D	0.48	0.83	0.65	0.29	1.02	0.82
Skin penetration (\(\log \ k_p\))	PUBCHEM	0.73	0.60	0.48	0.75	0.56	0.43
Human serum albumin	AP2D	0.71	0.33	0.23	0.69	0.39	0.26
Human placenta barrier	KR	0.41	0.24	0.20	0.24	0.32	0.22
Cancer potency in mouse (\(TD_{50}\))	AT2D	0.33	0.98	0.75	0.27	0.96	0.72
Cancer potency in rat (\(TD_{50}\))	AT2D	0.41	1.08	0.83	0.35	1.14	0.87
Steady state volume distribution (\(VD_{ss}\))	ASP	0.58	0.44	0.29	0.45	0.51	0.32
Distribution coefficient (\(\log \ D\))	PUBCHEM	0.76	0.73	0.53	0.77	0.71	0.50
Fraction unbound in human plasma	PUBCHEM	0.60	0.46	0.35	0.63	0.44	0.34
Fraction unbound in the brain	PUBCHEM	0.48	0.58	0.46	0.56	0.56	0.45
Human liver microsomal clearance	KR	0.51	1.08	0.80	0.56	1.05	0.79
Mouse liver microsomal clearance	AT2D	0.52	1.21	0.92	0.53	1.16	0.88
Rat liver microsomal clearance	KR	0.64	1.08	0.83	0.67	1.01	0.76
CACO-2 permeability	FCFP4	0.44	0.68	0.46	0.42	0.69	0.46
\({pK}_a\)	ECFP2	0.71	1.85	1.15	0.74	1.78	1.11
MDCK cell line permeability	ECFP4	0.62	0.61	0.44	0.68	0.56	0.39
Human renal clearance	MACCS	0.25	0.54	0.43	0.27	0.53	0.42
Hemolytic toxicity (\(\log \ {HD}_{50}\))	ASP	0.68	0.47	0.35	0.68	0.44	0.34

The values reported are the squared correlation (\(R^2\)), RMSE and MAE (average of 3 independent runs) for the calibration/validation sets

ISSN: 1758-2946