Modelling physico-chemical properties of halogenated benzenes: QSAR optimisation through variables selection

Area of Science:

• * Quantitative Structure-Activity Relationship (QSAR) modeling
• * Computational Chemistry
• * Environmental Chemistry

Background:

• * Halogenated benzenes are prevalent in environmental and industrial applications.
• * Accurate prediction of physico-chemical properties is crucial for risk assessment and process design.
• * Existing methods for determining properties like boiling point, melting point, flash point, and density can be resource-intensive.

Purpose of the Study:

• * To develop and validate multivariate QSAR models for predicting key physico-chemical properties of halogenated benzenes.
• * To utilize topological indices for comprehensive molecular characterization.
• * To establish reliable predictive tools for environmental and chemical applications.

Main Methods:

• * Development of multivariate QSAR models using 90 topological indices derived from chemical formulas.
• * Calibration of models on training sets selected via D-optimal design.
• * Optimization using statistical variable selection procedures and external validation for predictive capacity assessment.

Main Results:

• * Developed models demonstrated strong predictive accuracy for Melting Point (MP), Boiling Point (BP), Density (D), and Flash Point (FP).
• * Standard Deviation of Errors of Predictions (SDEP) values were ±21°C for MP, ±13°C for BP, ±0.14 g/cm³ for D, and ±17°C for FP.
• * External validation confirmed model reliability with experimental SDEP values of ±30°C for MP, ±16°C for BP, ±0.18 g/cm³ for D, and ±16°C for FP.

Conclusions:

• * The developed QSAR models provide a valuable and efficient tool for estimating essential physico-chemical properties of halogenated benzenes.
• * These models can significantly aid in environmental risk assessment and chemical process optimization.
• * The study highlights the utility of topological indices in predictive modeling for chemical compounds.