Development of molecular-solvent interaction descriptors to improve accuracy of Quantitative Structure Retention Relationship models

State of the art predictions of chromatographic retention rely on commercially available molecular descriptors such as VolSurf+3D, Dragon, MOE. These descriptors can be used to evaluate molecular structure-activity or structure-property relationships, as well as for similarity analysis and high-throughput screening of molecule databases. The algorithms that are used to calculate these descriptors usually do not take the composition of chromatographic mobile phase into account. For example, chromatographic mobile phase often consists of buffer/organic solvent mixtures which affect the ionic state of analytes and enhance or suppress specific interactions with the stationary phase. Although hydrophobic interactions are the primary retention mechanism, it is often the secondary interactions which determine the elution order of structurally similar compounds. The objectives of this proposal include the development of novel molecular descriptors that outperform currently available options by virtue of:
a) considering a medium more comparable with the chromatographic mobile phase
b) taking into consideration ion pairing with aqueous buffer components
c) differentiating between isomeric compounds e.g. positional isomers or diastereomers
Solvation descriptors will be developed using the integral equation theory of molecular liquids (IETML). IETML is a solvation model from statistical mechanics that provides a wealth of accurate solution-phase data, and is fast enough for high-throughput screening. The method implicitly differentiates between isomeric compounds, and is applicable to a wide-range of pure/mixed solvent systems, including those containing ionic species. It has been used successfully to compute solvation free energies, permeabilities, solubilities, partition coefficients, and binding free energies of druglike molecules [Chem. Rev., 2015, 115, 6312]. Solvation free energy density functions and thermodynamic parameters will be computed for each analyte for a matrix of different solvent compositions and temperatures. These solvation descriptors will be used to train novel Quantitative Structure Retention Relationship models (QSRR). We will follow a procedure similar to one used to predict Caco-2 permeability [Mol. Pharmaceutics, 2015, 12, 3420], but adapt it to chromatographic solvent systems and environmental conditions. All software packages that are developed by the University will be provided to Pfizer at the end of the project.