Quantitative structure-activity relationship for the partition coefficient of hydrophobic compounds between silicone oil and air.

The silicon oil-air partition coefficients (KSiO/A) of hydrophobic compounds are vital parameters for applying silicone oil as non-aqueous-phase liquid in partitioning bioreactors. Due to the limited number of KSiO/A values determined by experiment for hydrophobic compounds, there is an urgent need to model the KSiO/A values for unknown chemicals. In the present study, we developed a universal quantitative structure-activity relationship (QSAR) model using a sequential approach with macro-constitutional and micromolecular descriptors for silicone oil-air partition coefficients (KSiO/A) of hydrophobic compounds with large structural variance. The geometry optimization and vibrational frequencies of each chemical were calculated using the hybrid density functional theory at the B3LYP/6-311G** level. Several quantum chemical parameters that reflect various intermolecular interactions as well as hydrophobicity were selected to develop QSAR model. The result indicates that a regression model derived from logKSiO/A, the number of non-hydrogen atoms (#nonHatoms) and energy gap of ELUMO and EHOMO (ELUMO-EHOMO) could explain the partitioning mechanism of hydrophobic compounds between silicone oil and air. The correlation coefficient R2 of the model is 0.922, and the internal and external validation coefficient, Q2LOO and Q2ext , are 0.91 and 0.89 respectively, implying that the model has satisfactory goodness-of-fit, robustness, and predictive ability and thus provides a robust predictive tool to estimate the logKSiO/A values for chemicals in application domain. The applicability domain of the model was visualized by the Williams plot.