Modeling and predicting the solute polarity parameter in reversed-phase liquid chromatography using quantitative...

Area of Science:

• Computational Chemistry
• Analytical Chemistry
• Cheminformatics

Background:

• Quantitative structure-property relationships (QSPR) are crucial for predicting chemical compound properties.
• Accurate prediction of polarity is essential for reversed-phase liquid chromatography (RPLC) method development.
• Developing robust models for polarity prediction aids in understanding and optimizing chromatographic separations.

Purpose of the Study:

• To develop and compare QSPR models for predicting the polarity parameter of organic compounds in acetonitrile.
• To evaluate the performance of the enhanced replacement method (ERM) and support vector machine (SVM) regression for polarity prediction.
• To identify the most reliable modeling approach for RPLC applications.

Main Methods:

• Calculation of molecular descriptors solely from chemical structures.
• Application of enhanced replacement method (ERM) regression for QSPR modeling.
• Utilization of support vector machine (SVM) regression for QSPR modeling.
• Validation of models using experimental polarity parameter data for 146 organic compounds.

Main Results:

• Both ERM and SVM models were built using molecular descriptors.
• Correlation coefficients (R) for the test set were 0.970 for ERM and 0.993 for SVM.
• The SVM model demonstrated a higher correlation coefficient, indicating superior prediction accuracy.
• SVM exhibited better prediction performance compared to the ERM.

Conclusions:

• Support vector machine regression is a highly reliable method for predicting the polarity parameter of organic compounds in RPLC.
• The developed SVM model offers excellent prediction performance, outperforming the enhanced replacement method.
• This QSPR approach using SVM can significantly aid in chromatographic method optimization and compound characterization.