(717b) Predicting the Free Energy of Solvation: A Hybrid QSPR Model for Organic Solute/Solvent Pairs | AIChE

(717b) Predicting the Free Energy of Solvation: A Hybrid QSPR Model for Organic Solute/Solvent Pairs


Adjiman, C. S. - Presenter, Imperial College London
Garcia-Munoz, S., Eli Lilly and Company
Luciani, C., Eli Lilly and Company
Galindo, A., Imperial College London
The free energy of solvation is a useful quantity in the journey of a pharmaceutical molecule from discovery through to commercial process development and product design. In this paper, we review some of the roles played by this property across the molecule’s life cycle, including in the early assessment of the quantities such as the octanol/water partition coefficient or on the assessment of solvent effects on reaction energetics for screening purposes. Initial efforts to propose quantitative structure property relationships (QSPR) and predict the Gibbs free energy of solvation for a wide range of solute/solvent pairs are presented here. They include the use of partial least square (PLS) and multivariate linear regression (MLR) methods. A dataset consisting of 303 solutes in 215 solvents with total number of 1800 Gibbs free energy data points has been built. The proposed models combine experimental descriptors with quantum mechanical (QM) descriptors. For the solvents, experimental descriptors are proposed, as these provide a good description of the bulk properties of solvents and the range of solvents considered is usually relatively narrow; twelve of these descriptors are used in the proposed model (surface tension, dipole moment, octanol-water partition coefficient, dielectric constant, refractive index, boiling point, molecular weight, enthalpy of vaporization, molar volume, critical temperature, critical pressure, and critical volume). For the solutes, on the other hand, QM descriptors are used, making it possible to model solutes that have not been synthesized or that are short-lived/unstable; nine QM solute descriptors are proposed (polarizability, energy of the highest occupied molecular orbital, energy of the lowest unoccupied molecular orbital, dipole moment, most positive partial charge on hydrogen atom in molecule, most negative atomic partial charge in molecule, molecular van der Waals volume, total energy at 0 K, and entropy at 298 K). The accuracy of these initial models is found to be acceptable for fast calculation of the Gibbs free energy of solvation of a wide range of solutes in different solvents and to compare favourably with existing methods. Building on these encouraging results further work to explore whether this level of accuracy is suitable for a range of activities, including screening and process design will also be of interest.