EMPIRICAL PROCEDURE THAT USES MOLECULAR-STRUCTURE TO PREDICT ENANTIOSELECTIVITY OF CHIRAL STATIONARY PHASES

A total of 121 racemic compounds were separated in the normal-phase mode on a (S)-(1-naphthylethyl)carbamoylated beta-cyclodextrin (S-NEC-beta-CD) bonded phase and 74 on the R equivalent (R-NEC) chiral stationary phase (CSP). All compounds are of the type that have four substituents on a stereogenic center, rather than an "axis of chirality". It is shown that the binary solvent pair used as the mobile phase has a significant influence on chiral recognition. However, the proportions of the components of a specific pair have little effect. From the results, the individual contributions to chiral recognition by these CSPs were estimated for 81 different substituents of the stereogenic center. Varying the arrangement of these 81 substituents could produce over 1.6 million compounds. Hydrogen was chosen as the reference substituent and was assigned a 0 cal/mol free energy. The chiral recognition increased when sp2-hybridized carbons were connected to the stereogenic center. Conversely, sp3-hybridized carbons decreased the enantioselectivity. Amido groups increased the chiral recognition, especially when associated with pi-acid (3,5-dinitrobenzoyl) or pi-basic (naphthyl) groups. This approach does not allow one to know which enantiomer elutes first. However, the "substituent energy" list for chiral compounds can be used to obtain an estimated value for the enantioselectivity of a compound by adding the energy contributions of the four substituents connected to the stereogenic center. In this way one can predict a priori whether or not a compound will separate on a CSP and estimate its separation factor (alpha). Theoretically, this approach can be used for most CSPs, provided a sufficient data base is generated on them.