Linear Regression Based QSPR Models for the Prediction of the Retention Mechanism of Some Nitrogen Containing Heterocycles
Yulia Polyakova, Long Mei Jin, Kyung Ho Row · Journal of Liquid Chromatography & Related Technologies · 2006
This study evaluates retention factors of 29 nitrogen containing heterocycles using QSPR models in liquid chromatography. Some structure properties, such as the molecular connectivity indices (0χ∼5χ), Wiener index (W), Kier flexibility index (φ), Harary index (H), Balaban indices (JX∼JY), and Zagreb indices (M 1∼M 2) were obtained by theoretical molecular descriptors derived from information of chemical structures of substances. The relationship between the chromatographic retention factors and the structure descriptors were predicted using a mathematical method, regression analysis. All substances were divided into five groups based on similar structures with functional groups. For each group, the simple linear regression between each structural index and the retention factor showed good regression coefficients. The linear regression between retention factor and various orders of connectivity indices showed good regression coefficient (r 2=0.8∼1) and, with multiple structural properties, showed r 2=1. This study has demonstrated the successful linear regression approaches to prediction of the retention factors and some molecular descriptors of the substances.