HobPre: accurate prediction of human oral bioavailability for small molecules

Table 4 Comparison with other prediction models on test set 1 when the cutoff is 50%

Model	Data set size	Method	ACC (test)	AUC (test)	Cut-off value
Current study	1588	RF	0.793	0.830	F = 50%
Falcón-Cano et al. [8]	1448	CART, MLP, NB, GBT, SVM	0.783^a	0.800^a	F = 50%
admetSAR [9]	995	RF	0.697^a	0.752^a	logK(%F)^b = 0 (F = 50%)
Kim et al. [10]	995	RF, SVM-consensus CTG	0.76^a	NA	F = 50%

CART classification and regression trees; MLP multilayer perceptron; NB naive Bayes; GBT gradient boosted trees; SVM support vector machines
^aTaken from respective references
^b\(\mathrm{log}K\left(\mathrm{\%}F\right)=\mathrm{log}\left(\frac{\mathrm{\%}F}{100-\mathrm{\%}F}\right)\)

ISSN: 1758-2946