QSPR modeling of thermal conductivity detection response factors for diverse organic compound

Abstract

Quantitative structure property relationships (QSPRs) were applied to predict thermal conductivity detection response factors for diverse organic compound. One tested model is based on structural descriptors from molecular modeling. To quantitatively characterize the structure of analytes, the following four structural descriptors are employed: molecular weight (MW), sum of bond length (SBL), molecular polarizability effect index (MPEI), the product of molecular polarizability effect index and the pauling electronegativity ( χ p) of O-, N-atoms (MPEI × χ p). The last three descriptors were developed in our laboratory. Models of relationships between molecular structure and response factors (RFs) are constructed by means of multiple linear regressions (MLR). The high correlation of response factors with molecular descriptors was obtained. Additional validation was performed on an external data set consisting of 20 diverse organic compounds not involved in the deduction of the correlation equation from the main data set. Compared with an earlier model for the prediction of these compounds, our model exhibits slightly improved performance, and the selected molecular descriptors have explicit physicochemical meaning and easy to calculate. Furthermore, a whole number of descriptors were calculated with Dragon software and a subset of calculated descriptors was selected from Dragon descriptors with a forward stepwise MLR method which gives a similar superior prediction of the response factors as our model. The developed model in our work supports the identification and quantitation of substances by GC or GC-MS in cases response factors for candidate structures are not available.