Polarity parameters put in place by QSPR

Skip to Navigation


  • Published: Feb 15, 2018
  • Author: Ryan De Vooght-Johnson
  • Channels: Laboratory Informatics / Chemometrics & Informatics
thumbnail image: Polarity parameters put in place by QSPR

Solute polarity parameters need to be calculated

The prediction of accurate relative retention times for particular HPLC systems can save the time and expense of running many trials during method development. The aim of such work is to allow the prediction of retention times from the structures of the molecules of interest, an example of a quantitative structure–property relationship (QSPR).

The Azad University researchers examined two common techniques for calculating solute polarity parameters: support vector machine (SVM) and enhanced replacement method (ERM). The solute parameter (p), which was described by Bosch et al. in 1994, can be used to calculate retention times for a particular mobile phase and column. The Azad University researchers used an existing data set containing a wide range of organic compounds, 146 in all, where experimental solute polarity parameters had previously been determined under the same conditions using reversed-phase HPLC with aqueous acetonitrile as the solvent. The compounds varied from 4-hydroxybenzamide, which was the most polar compound, having a p value of 0.77, to n-butylbenzene, which was the least polar, with a p value of 6.13.

Support vector machine and enhanced replacement methods compared

All the chemical structures in the database were drawn using HyperChem version 7.0. Molecular descriptors of the compounds were then obtained using CODESSA version 2.7.2 software. The initial 214 descriptors were employed as inputs for the ERM, which was run using MATLAB version 7.0. The ERM is a technique for computing the optimum descriptors from a large set, giving a suitable regression equation. A five-variable (i.e. five-descriptor) optimum model was obtained, in which the solute parameter, p, was described by a five-variable regression equation. The five descriptors used were (in order of decreasing importance with regard to affecting p): qmax H, the maximum partial charge for an H atom; FPSA-3, a descriptor of charged partial surface area; qmin, the minimum net atomic charge (a descriptor related to charge distribution); pC, the maximum bond order for a carbon atom in the molecule; and V, the molecular volume.

It was decided to use the same five descriptors found for the ERM regression equation in the SVM calculations since it can be a problem deciding which descriptors to use for the latter technique. The SVM process was carried out using MATLAB version 7.0. SVM is a supervised learning model: in this case the compounds were divided into a training set (70% of the compounds) and a test set (30%). The SVM calculations involved the use of a so-called ‘kernel’ function, which determined the distribution of the samples in the mapping space.

Both the ERM and the SVM technique generally gave good predictions of p, comparable with those previously determined by experiment. Correlation coefficients were determined by comparing calculated and experimental values. The ERM gave a correlation coefficient, R, of 0.979 for the training set and 0.970 for the test set, while the SVM calculation gave R values of 0.997 and 0.993, respectively, showing that the latter method was superior.

SVM accurately predicts solute polarity parameters

The use of the SVM technique allowed solute polarity parameters, and hence retention times, to be determined with a high degree of accuracy. It would be useful if this method could be extended to other solvent systems, apart from aqueous acetonitrile, so that the best solvent system for separating particular compounds could then be determined. QSPR tools, such as SVM and the ERM, offer the potential to make analytical method development quicker and less costly.

Related Links

Journal of Separation Science, 2017, 40, 4495-4502. Golmohammadi et al. Modeling and predicting the solute polarity parameter in reversed-phase liquid chromatography using quantitative structure–property relationship approaches.

Analytical Chimica Acta, 1994, 299, 219-229. Bosch et al.

Wikipedia, Support Vector Machine

Article by Ryan De Vooght-Johnson

The views represented in this article are solely those of the author and do not necessarily represent those of John Wiley and Sons, Ltd.

Follow us on Twitter!

Social Links

Share This Links

Bookmark and Share


Suppliers Selection
Societies Selection

Banner Ad

Click here to see
all job opportunities

Copyright Information

Interested in separation science? Visit our sister site separationsNOW.com

Copyright © 2019 John Wiley & Sons, Inc. All Rights Reserved