Molecular modeling and prediction accuracy in Quantitative Structure-Retention Relationship calculations for chromatography

Quantitative Structure-Retention Relationship (QSRR) methodology is a useful tool in chromatography of all kinds, allowing the prediction of analyte retention time and providing insight into the mechanisms of separation. The prediction of retention is useful in reducing method development time and identifying analytes in Non-Targeted Analysis. The varying methods used for geometry optimization, descriptor calculation, feature selection, and model generation in many different QSRR settings are investigated and compared. It is found that the method of geometry optimization and descriptor selection is of less importance than the chromatographic similarity of compounds in the training sets used for model building in order to reduce the error of the model.