Determination of dermal absorption QSAR/QSPRs by brute force regression: multiparameter model development with...

Accurate dermal quantitative structure-activity/property relationship (QSAR/QSPR) models are needed to predict percutaneous absorption of environmental contaminants. The Molsuite 2000 chemistry modeling software (ChemSW, Fairfield, CA) was used to model the Flynn component of the Kirchner in vitro human skin permeability coefficient (K(p)) data. This Kirchner- derived Flynn (K/F) database was updated to include recent literature data quality recommendations. A K/F data subset consisting of nondrug compounds was used to further optimize the developed QSPR models. The statistical fit of the models was excellent with r(2) values up to.89 for a three-descriptor parameter K/F database model and up to.96 for a four- parameter model of the data subset. A one-parameter transform model using only logarithm octanol-water partition coefficient (log K(o/w)) was also developed for the data subset (r(2) =.87). Molecular volume (MV) descriptors were not shown to be superior to molecular weight in conventional two-parameter models with log K(o/w) but may be superior in multiparameter models. A previously nonreported descriptor, surface tension in water (STW), was found to provide optimal multiparameter models. The developed models passed PRESS cross-validation and could be useful for predicting environmental systemic dermal exposure.