Human effective intestinal membrane permeability (Peff) is one of the two important indicators for drug classification according to the Biopharmaceutical Classification System (BCS), and contributes greatly to the performance of oral drug absorption. Here, a structure-based in silico predictive model of Peff was developed successfully to facilitate in silico BCS classification in the early stage of drug discovery, even before the compound was synthesized. The quantitative structure-Peff relationship for 30 drugs was constructed based on seven structural parameters. Then the model was built by the multiple linear regression method and internally validated by the residual analysis, the normal probability-probability plot and the Williams plot. For the entire data set, the R² and adjusted R² values were 0.782 and 0.712, respectively. The results indicated that the fitted model was robust, stable and satisfied all the prerequisites of the regression models. As for the 102 tested drugs, the predicted Peff values had a good correlation with the experimental human absorbed fraction (Fa). This model was also used to perform high/low Peff classification for 57 drugs that have been classified according to the BCS, and 72% of drugs could be classified correctly, indicating that the developed model can be used for rapid BCS classification in the early stages of drug discovery.
Keywords: BCS; QSPR; computational ADME; human effective intestinal membrane permeability; multivariate analysis.
Copyright © 2013 John Wiley & Sons, Ltd.