Comparative molecular field analysis combined with physicochemical parameters for prediction of polydimethylsiloxane membrane flux in isopropanol.

Comparative molecular field analysis (CoMFA) combined with various physicochemical parameters were used to develop three-dimensional quantitative structure-transportability relationships (3-D QSTR) to predict membrane flux for 108 aromatic and heteroaromatic compounds through polydimethylsiloxane (PDMS) membranes in isopropyl alcohol (IPA). Sybyl, a comprehensive computational molecular modeling package, was used to analyze the data. Optimized molecular models were selected using molecular modeling techniques. Partial least-squares (PLS) regression combined with crossvalidation or bootstrapping was used as the statistical method to establish the predictive models. Prediction was good for the steady-state flux using both steric and electrostatic field descriptors combined with a functional group classification technique. Predictive ability was substantially increased in a model using CoMFA descriptors along with log mole fraction solubility for the penetrants in isopropanol, a hydrophobic term, fchex, which is used to estimate the partition coefficient between cyclohexane and water, and the addition of an intramolecular hydrogen bonding (IHB) term. The crossvalidated r2 and the conventional r2 for this model were 0.951 and 0.973, respectively. Excellent predictions are demonstrated for the membrane flux of the compounds both inside and outside the data domain.