Table 1.
Statistic parameters correlation coefficient (r2), maximum residual (ΔMax), mean absolute error (MAE), standard deviation of residual (s), RMSE, and cross-validation coefficient evaluated by Hypo A, Hypo B, Hypo C, and PhE/SVM in the training set.
Table 2.
Statistic parameters correlation coefficients ,
, and
, concordance correlation coefficient (
), maximum residual (ΔMax), mean absolute error (MAE), standard deviation of residual (s), and RMSE evaluated by Hypo A, Hypo B, Hypo C, and PhE/SVM in the test set.
Figure 1.
Pharmacophore models in the ensemble.
Generated pharmacophore models (A) Hypo A, (B) Hypo B, and (C) Hypo C, consisting of hydrogen-bond acceptor (green), hydrophobic (light blue), and ring aromatic (orange) chemical features. The interfeature distances and angles among features, depicted in white, are measured in Ångstroms and degrees, respectively.
Figure 2.
Superposed pharmacophore models.
Superposition of three pharmacophore models Hypo A, Hypo B, and Hypo C, denoted in red, blue, and green, respectively.
Figure 3.
Observed vs. predicted pIC50 values in the training set.
Observed pIC50 vs. the pIC50 predicted by Hypo A, Hypo B, Hypo C, and PhE/SVM for those molecules in the training set. The solid line, dashed lines, and dotted lines correspond to the PhE/SVM regression of the data, 95% confidence interval for the PhE/SVM regression, and 95% confidence interval for the prediction, respectively.
Figure 4.
Superposition of pharmacophore models and 22.
Pharmacophore models (A) Hypo A, (B) Hypo B, and (C) Hypo C fitted to 22 and (D) overlay of these three models, which are color-coded by red, blue, and green, respectively. The chemical features are described in Figure 1.
Figure 5.
Observed vs. predicted pIC50 values in the test set.
Observed pIC50 vs. the pIC50 predicted by Hypo A, Hypo B, Hypo C, and PhE/SVM for those molecules in the test set. The solid line, dashed lines, and dotted lines correspond to the PhE/SVM regression of the data, 95% confidence interval for the PhE/SVM regression, and 95% confidence interval for the prediction, respectively.
Table 3.
Optimal runtime parameters for the SVM model.
Figure 6.
Sample distribution in the chemical space.
Molecular distribution for those samples in the training set (blue circle), test set (green triangle), and outlier set (red square) in the chemical space spanned by three principal components.
Figure 7.
Observed vs. predicted pIC50 values in the outlier set.
Observed pIC50 vs. the pIC50 predicted by Hypo A, Hypo B, Hypo C, and PhE/SVM for those molecules in the outlier set. The solid line, dashed lines, and dotted lines correspond to the PhE/SVM regression of the data, 95% confidence interval for the PhE/SVM regression, and 95% confidence interval for the prediction, respectively.
Table 4.
Statistic parameters correlation coefficients ,
, and
, concordance correlation coefficient (
), maximum residual (ΔMax), mean absolute error (MAE), standard deviation of residual (s), and RMSE evaluated by PhE/SVM in the outlier set.
Figure 8.
Residual vs. predicted pIC50 values.
Residual vs. the pIC50 predicted by PhE/SVM in the training set (filled circles), test set (open triangles), and outlier set (gray squares).
Table 5.
Validation verification of PhE/SVM based on prediction performance of those molecules in the training set, test set, and outlier set.
Figure 9.
Observed fold increase vs. observed IC50.
Experimental fold increase of BCRP inhibition measured in Saos-2 cells vs. observed IC50 measured in MCF-7 MX cells.
Figure 10.
Observed fold increase vs. predicted IC50.
Experimental fold increase vs. the IC50 predicted by PhE/SVM for those HIV protease inhibitors.
Table 6.
Summary of developed BCRP inhibition qualitative and quantitative pharmacophore hypotheses.
Figure 11.
Model proposed by Matsson et al. and excerpted model of this study.
Geometrical relationships in the pharmacophore models (A) proposed by Matsson et al. and (B) excerpted from the PhE in this study. The interfeature distances are measured in Ångstroms.
Figure 12.
Model proposed by Sim et al. and excerpted model of this study.
Geometrical relationships in the pharmacophore models (A) proposed by Sim et al. and (B) excerpted from the PhE in this study. The interfeature distances are measured in Ångstroms.