The therapeutic efficacy of an orally administered drug is dictated not only by its pharmacological properties such as potency and selectivity, but also its pharmacokinetic properties such as its access to the site of activity. Thorough evaluation of the physicochemical and biological barriers to drug delivery is essential to the selection and successful development of drug candidates. We have demonstrated previously that cellular permeability, as a primary component of drug delivery, is principally dependent upon the desolvation potential of the polar functionalities in the molecule and, secondarily, upon the solute lipophilicity [Conradi, R.A., Hilgers, A.R., Ho, N.F.H., Burton, P.S. (1992) The influence of peptide structure on transport across Caco-2 cells. II. Peptide bond modification which results in improved permeability. Pharm. Res.9, 473-479 ]. Increasingly sophisticated computational methods are becoming available for describing molecular structural features proposed to correlate with such molecular physicochemical determinants of permeability. Herein we examine the relationships of various computationally derived molecular geometric descriptors for a set of peptides and peptidomimetics, in the context of experimentally measured hydrogen-bond potentials and lipophilicities, with their cellular permeabilities. These descriptors include molecular volume, polar and non-polar surface areas and projected molecular cross-sectional areas. Particular attention is paid to the roles of solvation treatments and other computational factors in descriptor generation, deconvolution of cellular transport mechanisms and statistical analyses of the resulting data for the development of valid, structure-based and mechanistically meaningful models of cellular permeability. No significant correlation of cellular permeability with computed descriptors was found. This was primarily because of our inability to identify surrogates for hydrogen-bond desolvation potential for the solutes from among these descriptors.