QSPR models for n-octanol/water partition coefficient and enthalpy of vaporization using CDFT and information theory-based descriptors

The quantitative structure-property relationship (QSPR) technique is used to gauge the n-octanol/water partition coefficient (log K_OW) and enthalpy of vaporization (∆_vapH_m) of 133 Polychlorinated Biphenyls (PCBs) using conceptual density functional theory (CDFT)-based global reactivity and information-theory (IT) based parameters. Regression models are established using linear and multi-linear relationships to correlate the observed physicochemical properties of PCBs with the predicted ones. The study explored the significance of CDFT and IT descriptors, and based on the calculation of Pearson correlation coefficient values, the selection of suitable descriptors is made for successful QSPR models of selected PCBs. It is found that some of the CDFT parameters are highly correlated with the IT parameters, as suggested by their high Pearson correlation coefficient values for PCB systems. The regression model generated using the descriptors I_G, g₁, g₂, EA, η for predicting log K_OW and I_F, g₃, η, S_S, S_GBP for predicting ∆_vapH_m gives R² value of 0.9342 and 0.8662, respectively, for the selected 133 PCB congeners. Furthermore, to verify the descriptor selection, a machine learning approach is also used to develop QSPR models in this study.

Graphical Abstract

QSPR modelling using CDFT and information theory-based descriptors for predicting n-octanol/water partition coefficient and enthalpy of vaporization for the selected PCBs