QSPR correlation and predictions of GC retention indexes for methyl-branched hydrocarbons produced by insects

Abstract

A successful interpretation of the complex manner by which the GC retention indexes of methylalkanes produced by insects are related to chemical structure was achieved using the quantitative structure-property relationship (QSPR) method. A general QSPR model including mainly topological descriptors was obtained for 178 data points. The error of the model is similar to the experimental error. The model was supported by (i) leave-one-out cross validation and (ii) division into three sets and prediction of each set from the other two. As a further test of the utility of the model, retention indexes were successfully predicted for an external set of 30 methyl-branched hydrocarbons not involved in the deduction of the correction equation from the main data set. General trends of the structural variation of compounds in any given range of retention index are discussed. The average error was 4.6 overall and 4.3 for the 165 compounds remaining after leaving out small monomethyl alkanes.