Atomic-Level Topological Indices for Prediction of the Infinite Dilution Activity Coefficients of Oxo Compounds in Water

Abstract

The atomic-level AI topological indices and the modified Xu (mXu) index were utilized for quantitative structure–property relationship (QSPR) modeling of the infinite dilution activity coefficients of 108 oxo compounds in water at 298.15 K. Stepwise multiple linear regression (SMLR) analysis using the topological descriptors resulted in a model with R2 = $${\mathrm{R}}_{\mathrm{a}\mathrm{d}\mathrm{j}}^{2}$$= 0.9904, SE = 0.3769, F = 1267.1 and an average relative error of 4.89%. The selected descriptors were then used to develop an artificial neural network (ANN) model for the activity coefficients. Findings of the study indicated that a 7–8-1 ANN trained by Levenberg–Marquardt algorithm results in the improved predictions, especially in view of a decrease as large as 47.24% in the average relative error compared to the SMLR model. The AI indices with a total contribution of 81.43% showed the dominant role of the atomic groups of the oxo compounds in determination of their activity coefficients at infinite dilution in water.