A number of QSAR models for skin permeability have been proposed, but these models lack consistency due to interspecies and interlaboratory differences. This study was initiated to extract an essential QSAR from the multiplicity of data sets of skin permeability by using a novel statistical approach. Ten data sets were collected from the literature, which include a total of 111 permeability coefficients in human, hairless mouse, or hairless rat skin for 94 structurally diverse compounds. Following a Potts and Guy's approach, the octanol/water partition coefficient and molecular weight were chosen as molecular descriptors. All of the data sets were analyzed simultaneously, assuming that all of the sets share a latent, common factor as far as the structure/permeability relationship is concerned. Despite the fact that the degree-of-freedom for the present analysis was limited compared with that for individual regression analyses, the determination coefficients (R2) were high enough for all the 10 data sets, with an average R2 of 0.815 (average R2 = 0.825 for individual analyses). Thus, skin permeability of compounds can be well explained from the log P and M.W., where the ratio of the contribution to skin permeability was approximately 1:1.
Copyright 2003 Wiley-Liss, Inc. and the American Pharmacists Association