Predicting Caco-2/MDCK intrinsic membrane permeability from HDM-PAMPA-derived hexadecane/water partition coefficients

Reliable membrane permeability data are essential in early drug development. Therefore, there is a strong need for robust experimental high-throughput screening methods, or ideally, accurate predictive tools, to assess membrane permeability. In a previous study, we demonstrated that the solubility-diffusion model can successfully predict passive permeability across biological Caco-2 and MDCK membranes, provided accurate hexadecane/water partition coefficients (K_hex/w) are available. In this study, we investigated the HDM-PAMPA method for determining K_hex/w. We measured our own data (64 compounds) using this assay and compared the results with established methods such as black lipid membrane (BLM) experiments and classical two-phase systems. Our results show good agreement across methods, with both our data and literature values aligning closely. Using these experimentally determined K_hex/w values, we achieved accurate predictions of permeability in Caco-2 and MDCK cell membranes (RMSE = 0.8, n = 29) based on a previously calibrated equation. We further evaluated the in silico prediction of K_hex/w using the UFZ-LSER database and the software COSMOtherm. COSMOtherm performed nearly as well as experimental measurements (RMSE = 1.20, n = 29), while the LSER approach (RMSE = 1.63, n = 29) is best applied when experimental descriptors are available or as a complement to COSMOtherm. This work highlights the practical utility of K_hex/w in high-throughput permeability estimation, which can support efficient screening and prioritization of drug candidates in pharmaceutical research.