QSPR analysis of HPLC column capacity factors for a set of high-energy materials using electronic van der waals surface property descriptors computed by transferable atom equivalent method

Abstract

Listen

Translate article

The new transferable atom equivalent (TAE) method for rapid molecular electron density reconstruction was used to compute a set of molecular surface property descriptors. These descriptors were then used to construct HPLC column capacity factor PLS models for a series of high-energy materials. The new TAE-derived surface property indices are also available from ab initio or semiempirical wave functions, but the speed and accuracy of TAE reconstruction make it the method of choice for obtaining these indices. The new QSPR indices are based upon the extrema, distributions, and surface integrals of the electronic kinetic energy density, the Politzer average local ionization potential (pip), and the electrostatic potential, as well as the rates at which these properties change normal to the 0.002-e/au3 molecular surface. The distribution of the properties were recorded as surface histograms. While property extrema and surface integral averages proved to be descriptive, the most useful new indices were found to correspond to histogram bin data computed for K and G surface kinetic energy densities. All-subsets-regression modeling showed that when mixtures of traditional connectivity indices, theoretical linear solvation energy relations (TLSERs), and generalized interaction properties functions (GIPFs) were included with the new indices in the variable sets, the new indices were consistently involved in the best 2% of the capacity factor models. Peak retention time data from two different columns were examined (a Hypersil “CPS” cyanoalkylated column and a standard reverse-phase Hypersil “ODS” column) using a phosphate buffer mobile phase at pH 3.0. The result were compared to an earlier TLSER correlation analysis of the same data by Lowrey and Famini. The TAE-generated surface property descriptors were shown to provide superior PLS models for both sets of columns and conditions. © 1997 by John Wiley & Sons, Inc.