Table 1.

Overview of different HT-PBK modelling strategies and their predictive performances

		Best overall (including in vivo and in vitro benchmark values)	Best in silico-proprietary	Best in silico-free
Input parameter sources	Lipophilicity	Mean of Bayer LogD and LogMA	Mean of Bayer LogD and LogMA	ADMETLab 2.0: LogD
	Solubility	SimPlus: FaSSIF	SimPlus: FaSSIF	ADMETLab 2.0: Aqueous solubility
	pKa	ChemAxon	ChemAxon	None
	Partitioning method	PK-Sim	PK-Sim	PK-Sim
	Fraction unbound	In vitro measured	ADMETLab 2.0	ADMETLab 2.0
	Clearance	In vivo observed plasma clearance	ADMETLab 2.0: Plasma clearance	ADMETLab 2.0: Plasma clearance
	Intestinal permeability	ADMETLab 2.0: MDCK	ADMETLab 2.0: MDCK	ADMETLab 2.0: MDCK
Performance metrics	IV: median absolute Log2 error	0.860	1.065	1.273
	PO: median absolute Log2 error	1.245	1.595	1.780
	PO: median absolute Log2 Cmax predicted/observed	0.998	1.127	1.130
	PO: compounds with Cmax within ten-/five-/three-/twofold	86%/78%/69%/50%	87%/78%/65%/44%	89%/71%/62%/42%
	PO: underprediction of compound Cmax of more than ten-/five-/three-/twofold	1.9%/4.3%/8.1%/16%	0%/3.1%/8.1%/19%	3.1%/12%/16%/25%
	PO: median absolute Log2 AUC predicted/observed	0.870	1.285	1.358
	PO: compounds with AUC within ten-/five-/three-/twofold	92%/84%/75%/57%	84%/71%/58%/38%	84%/66%/54%/39%