Skip to main content
. Author manuscript; available in PMC: 2018 Jul 4.
Published in final edited form as: Eur J Med Chem. 2017 May 10;137:126–138. doi: 10.1016/j.ejmech.2017.05.026

Table 1.

Statistical characteristics of developed QSAR models estimated by 5-fold external CV.

Models CCR Kappa Se Sp Coverage
MACCS-GBM 0.73 0.46 0.76 0.70 0.71
AtomPairs-GBM 0.71 0.41 0.71 0.70 0.71
Morgan-GBM 0.76 0.51 0.77 0.74 0.68
FeatMorgan-GBM 0.74 0.47 0.76 0.71 0.66
Avalon-GBM 0.74 0.47 0.79 0.68 0.77
MACCS-RF 0.75 0.51 0.79 0.72 0.71
AtomPairs-RF 0.75 0.50 0.73 0.77 0.71
Morgan-RF 0.76 0.52 0.79 0.73 0.68
FeatMorgan-RF 0.75 0.50 0.70 0.80 0.66
Avalon-RF 0.74 0.49 0.76 0.73 0.77
MACCS-SVM 0.77 0.53 0.77 0.76 0.71
AtomPairs-SVM 0.74 0.48 0.74 0.74 0.71
Morgan-SVM 0.76 0.53 0.80 0.73 0.68
FeatMorgan-SVM 0.76 0.51 0.75 0.76 0.66
Avalon-SVM 0.73 0.46 0.72 0.74 0.77
Consensus* 0.77 0.53 0.79 0.74 1.00

GBM: Gradient Boosting Machine; SVM: Support Vector Machine; RF: Random Forest; CCR: correct classification rate; Kappa: Cohen’s kappa coefficient; Se: sensitivity; Sp: specificity.

*

Consensus model was developed by averaging the predictions of all 15 single models.