More than half of all pharmaceuticals are chiral compounds. Although the enantiomers of chiral compounds have the same chemical structure, they can exhibit marked differences in physiological activity; therefore, it is important to remove the enantiomer that exhibits undesirable physiological activity. Chromatographic separation of chiral enantiomers is one of the best available methods to obtain enantio-pure substances, but the optimization of the experimental conditions can be very time-consuming. Generally, there are several chiral stationary phases to choose from; the most commonly used ones are polymers that have an amylose or cellulose backbone, with various sidechains. These could be coated on the solid support or covalently bonded to it. Then there is the choice of the solvent system that constitutes the mobile phase, dictated by whether the CSP is coated or covalently bonded to the support, and also by the solubility of the drug. Furthermore, modifiers (acids or bases or buffers) may be added to the solvent in order to change or protect some functional groups on the drug to realize a better resolution between the enantiomers. One could try different CSPs and mobile phases using educated guesses based on experience, but the number of possible combinations is huge, and could easily run into hundreds of combinations. To reduce possible combinations of experimental approaches to chiral separations, we proposed a computational method that would provide the largest separation between the enantiomers of the drug in question. Chromatographic separation is a dynamic process, with the interactions between the drug and the chiral stationary phase mediated by the solvent, so no single property, such as the minimum interaction energy, could possibly describe and account for the ratio of residence times in the chromatographic column for the enantiomeric pair. We chose to use explicit-solvent fully atomistic molecular dynamic simulations, permitting all the interactions between the atoms constituting the polymeric chiral stationary phase, the solvent molecules and the drug molecule.
Using molecular dynamics simulations, we have developed a predictive molecular model to determine the experimental elution order of enantiomers and the experimental value of the separation factor. This predictive capability is beyond what HPLC separation experiments can provide. The latter can only provide separation factors and additional experiments are required to find which absolute configuration enantiomer was eluted first. Our model predicts which of the enantiomers elutes first. At the present stage of this work, the best correlation provides separation factors with a correlation coefficient of 0.80 and the elution order is predicted correctly for 90% of the cases. The metric which provides this level of prediction is the maximum value of the average lifetime of hydrogen bonds between the drug and the CSP, including all donor-acceptor pairs between the drug and the CSP in this overall measure.
In our next phase of work, we are investigating several other CSPs, not only different side chains on the amylose backbone, but also cellulose-based CSPs, investigate the effect of using modifiers (small concentrations of a compound that is added to the solvent system in order to improve the baseline separation and improve the separation factor for a given drug/CSP/solvent system combination), and investigate the currently widely-used supercritical media, such as supercritical CO2 .
