QSPR Analysis with Curvilinear Regression Modeling and Temperature-based Topological Indices

TL;DR

This paper develops QSPR models using temperature-based topological indices and curvilinear regression to predict thermodynamic properties of monocarboxylic acids, enhancing understanding of molecular structure-property relationships.

Contribution

It introduces novel temperature-based topological indices and applies curvilinear regression models to improve prediction accuracy of thermodynamic properties.

Findings

Curvilinear regression models outperform linear models.

Temperature-based topological indices effectively predict thermodynamic properties.

Models show high correlation with experimental data.

Abstract

Establishing quantitative correlations between various molecular properties and chemical structures is of great technological importance for environmental and medical aspects. These approaches are referred to as Quantitative Structure-Property Relationships (QSPR), which relate the physicochemical or thermodynamic properties of compounds to their structures. The main goal of QSPR studies is to find a mathematical relationship between the property of interest and several molecular descriptors derived from the structure of the molecule. Topological indices are the molecular descriptors that characterize the formation of chemical compounds and predict certain physicochemical properties. In this study, the QSPR models are designed using certain temperature-based topological indices such as the sum connectivity temperature index, product connectivity temperature index, F-temperature index, and symmetric division temperature index to predict the thermodynamic properties, such as enthalpies of formation ($\Delta H^{0}_{f}$ \hspace{1mm} liquid), enthalpies of combustion ($\Delta H^{0}_{C}$ \hspace{1mm} liquid), and enthalpies of vaporization ($\Delta H^{0}_{vap}$ \hspace{1mm} gas) of monocarboxylic acids ($C_2H_{4}O_{2}$ - $C_{20}H_{40}O_{2}$). The relationship analysis between thermodynamic properties and topological indices is done using linear, quadratic, and cubic equations of a curvilinear regression model. These regression models are then compared.