Partitioning of active agents between polar and nonpolar phases has a key role in the early stage of drug and drug-carrier design in the pharmaceuticals industry, as well as for separation of products in biosynthesis. In the present paper, the group-contribution Universal Quasi-Chemical Functional-Group Activity Coefficient (UNIFAC) and the a priori Conductor-like Screening Model for Real Solvents (COSMO-RS) models are applied to predict micelle/water partition coefficients. The models allow predictions based only on the molecular structure. The practical implementation of the models is examined by studying several homologous series of organic solutes in aqueous solutions of non-ionic (polyethoxy alcohols) and ionic surfactants (sodium dodecyl sulfate (SDS)). Good quantitative agreement with experimental data from the literature has been achieved. Factors that seem to be important in the calculation and to influence the prediction results are discussed. Among these are interfacial contribution and conformation analysis. Compared to UNIFAC, the COSMO-RS method opens up new perspectives, because ionic components, steric isomers, and inorganics can be modeled.

This work is encouraged by the growing interest and recent success of quantum mechanics-based methods in modeling of thermodynamic properties in the field of chemical engineering and life sciences. Among those, the COSMO-RS model has become one of the most popular methods to predict phase equilibria in complex bio-related systems. Recently, we have shown that the quality of predictions of n-octanol/water and micelle/water partition coefficients is improved when the weighted mixtures of conformers are used to represent molecules, thereby demanding that the conformation analysis is performed for each component of the system. In this paper, different methods for performing the conformational search are evaluated. Micelle/water partition coefficients of solutes from different homologous series in aqueous solutions of Triton X100 as well as the n-octanol/ water partitioning of three common drugs (aminopinicillanic acid, ampicillin, and penicillin G) are calculated and compared with experimental data. Conformational analysis is made by the HyperChem program for the molecules placed in vacuum as well as by using the molecular dynamics simulation in a solvent medium (n-octanol and water). It is demonstrated that molecular dynamics simulation is a promising tool to conduct conformational analysis. Along with the possibility of providing the conformers for large surfactant and pharmaceutical molecules, the method accounts for the solvent in a realistic manner. The results for micelle/water partition coefficients illustrate that n-octanol is a reasonable approximation for the "micelle-like" medium in molecular dynamics (MD) simulations. Predicted n-octanol/water partition coefficients of three penicillins are in a good agreement with literature data and values calculated by a common quantitative structure-activity relationship (QSAR).

In the present study the potential of two thermodynamic-based models (the group-contribution UNIFAC model and the a priori predictive COSMO-RS model) to predict solute partitioning in aqueous surfactant solutions is evaluated. In order to take into account the small size of micelles, the UNIFAC model was extended by the interfacial contribution based on the Gibbs-Thompson equation. The applicability of the approach was successfully proved for the partitioning of nonpolar solutes (toluene, p-xylene) in aqueous solutions of nonionic surfactants. The original COSMO-RS model underestimates the concentration of the solute in the aqueous phase in the case of micellar systems since it does not account for the small size of micelles. At the same time, this model gives quantitative results for the octanol/water partition coefficients of both solutes under study leading to the conclusion that the affinity of both solutes to a certain solvent is well described. Thus, extending the COSMO-RS by the interfacial term seems to be promising.