AB-INITIO PREDICTION OF HUMAN SKIN PERMEABILITY COEFFICIENTS

A general linear model was used to fit published human stratum corneum log permeability coefficient (log P) values for 91 diverse compounds to various parameters of their molecular structures. This enabled calculation of additive group contribution values which are useful in predicting log P values. Two sets of predictors are suggested based on the SMILES method of molecular structure description and on an 11-predictor set of empirically determined functional groups. The results were in general agreement with the prediction method based on octanol/water partition and molecular weight proposed by Potts and Guy (Pharm. Res., 9 (1992) 663-669). The method has obvious application in the simple estimation of permeability of compounds for which no data are published, and, where the predicted values for the three methods differ significantly from experimental values, could indicate compounds in which the permeation processs is abnormal.