Pieter De Gauquier

In this work, chiral molecular descriptors were defined using 2 distinct approaches: (1) scalar triple products of vectorial molecular properties, and (2) descriptors that attempt to quantify the amount of twist in the overall molecular shape. Because both approaches give rise to conformation dependence, descriptor values were averaged over a conformational ensemble obtained by Molecular Dynamics. In addition, a method is introduced that attempts to quantify the asymmetry of the distribution of the descriptor values over the conformational ensemble. The totality of the resulting descriptors were named “Ensemble Steric and Electrostatic Chirality (ESEC) descriptors”. A pilot validation study was performed by building Quantitative Structure-Enantioselectivity Relationships (QSER), i.e. mathematical models to predict the chromatographic separation of enantiomers, using a test set of 43 structurally diverse pharmaceuticals analyzed on a polysaccharide-based chiral stationary phase. The best linear regression model (7 descriptors) for the chiral separation (expressed as selectivity factor) featured a low leave-one-out cross validation error (0.0814), a well-predicted elution sequence of the separated enantiomers (21 out of 23 molecules) and a well-predicted α _RS for 27 out of 42 molecules. To the best of our knowledge, this is the first time that acceptable linear QSER models were obtained for chiral chromatographic separations of such a chemically diverse set of pharmaceuticals.