Atom-type-based AI topological descriptors for quantitative structure–retention index correlations of aldehydes and ketones

The gas chromatographic retention data (IR) of a mixed set of aldehydes and ketones on different polar stationary phases (HP-1, HP-50, DB-210, and HP-Innowax) are well correlated with the recently introduced Xu index and atom-type-based AI topological indices using the multiple linear regression (MLR) method (r>0.995). The results indicate that the molecular size makes a dominant contribution to IR, but atom types or groups also provide smaller and separate contributions. It is found that the carbonyl group makes a greater contribution to retention indices on polar columns than the –CH3, >CH–, and >C< groups due to stronger polar interactions between eluents and stationary phases. The results suggest that polar interactions between solute molecules and polar stationary phases become more and more important to retention indices with increasing the polarity of columns. Furthermore, branching of a molecule is also important to retention indices on different polar columns, particularly polar HP-210 and HP-Innowax columns. The models with topological indices are compared with those based on quantum-chemical descriptors and physicochemical properties, respectively. It is found that topological indices produce better correlations with Kováts retention indices than physicochemical properties for all four columns and also give better correlations than quantum-chemical descriptors for three out of four stationary phases. The results indicate the efficiency of these indices in the structure–retention index correlations of complex compounds with polar functional groups. The leave-one-out cross-validation demonstrates the final models to be statistically significant and reliable.