Introduction: Quantitative structure-retention relationships (QSRR) models were built for a data set consisting of 89 saturated esters and used to predict their gas chromatographic Kováts relative retention indices (GC-RRIs) on ten stationary phases of different polarity including SE-30, OV-7, DC-710, OV-25, 100% phenyl, DC-230, DC-530, XE-60, OV-225 and Silar-5CP [1,2]. Methods: In this investigation, AM1 semi-empirical molecular orbital method was applied to optimize the molecules geometry. The sets of 30 molecular connectivity indices (MCIs) were derived directly from the topological descriptors from DRAGON program [3]. Then, multiple linear regression (MLR) models with three optimal descriptors based on the whole data set were obtained to predict the RRIs of ester compounds on each stationary phase [4]. These descriptors that appear in the best MLR equations for the ten stationary phases are identical. Results and discussion: Excellent results were obtained employing stepwise multiple linear regressions and critically discussed using a variety of statistical parameters (R2 between 0.9398 and 0.9580, R2CV between 0.9341 and 0.954, Fisher ratio between 442.6 and 646.9). Furthermore, the QSRR models were validated using leave-one-out cross-validation (LOO), Y-randomization, external validation through an odd-even number and division of the entire data set into training and test sets. Conclusions: These confirm that the obtained regression models have a good internal and external-predictive power. The result of the study indicated the efficiency of the molecular connectivity approaches.
