New atom-type-based AI topological indices: Application to QSPR studies of aldehydes and ketones

Abstract

Multiple linear regression (MLR) analysis based on a combined use of the modified Xu index and the atom-type based AI indices is performed to construct quantitative structure-property models on several data sets of organic compounds including aliphatic aldehydes and/or ketones. For each of the physical properties (the normal boiling points, molar refractions, gas heat capacities at 25 degrees C, water solubility at 25 degrees C, and n-octanol/water partition coefficient at 25 degrees C), high quality QSPR models are obtained, particularly the decrease in the standard error is within the range of 23.6-75.9% relative to the linear models with the modified Xu index alone. For individual subsets containing only aldehydes or ketones, in the majority of cases the quality of the model can be further improved. The significant improvement verifies the efficiency of the present approach and also indicates the usefulness of these indices for application to a wide range of physical properties. The results indicate that the physical properties studied are dominated by molecular size but atom types have smaller influences, especially the oxygen atom seems to be most important due to intermolecular polar interactions. The final models are validated to be statistically reliable using the leave-one-out cross-validation and/or an external test set.