Electrotopological state index (ETSI) for atom types was used to describe the structures of 35 organophosphates and a quantitative linear relationship between the ETSI descriptors and gas chromatographic retention indices (RI) was developed using the variable selection and modeling based on prediction (VSMP). It was found that some main structural factors influencing the RI of organophosphates are 7 substructures such as ＝CH₂, ≡C—, aaC— (where “a” refers to a chemical bond in the aromatic ring), ＝O, O, Cl and Br, which were related to the molecular skeleton of organophosphates, substituent groups on phenyl ring, and alkyls binding to the bond of P—O. Three best 7-variable models, with the calibrated correlation coefficient of r＞0.99 and the validated correlation coefficient of q＞0.98 for three stationary phases, were built by multiple linear regression, which shows a good estimation ability and stability of models. A prediction power for the external samples was validated by the model built from the training set with 28 organophosphates.

Key words: electrotopological state index, organophosphate, quantitative structure-retention relationship, variable selection and modeling based on prediction