In Silico Prediction of Drug Permeability Across Buccal Mucosa

Abstract

To develop and validate a computational model capable of predicting buccal permeability based on various structural and physicochemical descriptors. Apparent permeability coefficients (K(p)) of 15 different drugs across porcine buccal mucosa were determined. Multiple linear regression (MLR) and maximum likelihood estimations (MLE) were used to develop the model based on a training set of 15 drugs with permeability as the response variable and the various descriptors as the predictor variables. The final model was validated with an external data set consisting of permeability values obtained from the literature. Drug permeabilities ranged from 30 x 10(-6) (nimesulide) to 3.3 x 10(-9) cm/s (furosemide). Regression analysis showed that 95% of the variability in permeability data can be explained by a model that includes molecular volume, distribution coefficient at pH 6.8, number of hydrogen bond donors, and number of rotatable bonds. Smaller molecular size, high lipophilicity, lower hydrogen bond capability and greater flexibility were important for permeability. The buccal model was found to have a good predictive capability. A simple model was developed and validated for predicting the buccal drug permeability. This model will be useful in assessing the feasibility of drugs for transbuccal delivery.