Fast approaches for molecular polarizability calculations

Area of Science:

• Computational Chemistry
• Molecular Modeling
• Cheminformatics

Background:

• Molecular polarizability quantifies electronic distortion by external fields, crucial for predicting molecular properties and biological activities.
• Accurate polarizability prediction is vital for applications like Quantitative Structure-Activity Relationship (QSAR) and Quantitative Structure-Property Relationship (QSPR) studies.

Purpose of the Study:

• Develop fast empirical models for predicting molecular polarizability.
• Compare two distinct modeling approaches: Slater's rules and additive hypotheses.
• Identify the most accurate models for broad application in computational studies.

Main Methods:

• Developed empirical models based on Slater's rules for effective atomic nuclear shielding constants (Model 1A).
• Constructed additive models using various atom type schemes, with the best employing 14 atom types (Model 2E).
• Evaluated model performance using Average Unsigned Error (AUE), Root-Mean Square Error (RMSE), and Average Percent Error (APE).

Main Results:

• Model 1A (Slater's rules) achieved AUE of 2.23 au, RMSE of 3.29 au, and APE of 2.77%.
• Model 2E (additive hypothesis with 14 atom types) demonstrated significantly higher accuracy with AUE of 0.99 au, RMSE of 1.48 au, and APE of 1.24%.
• Model 2E outperformed models based solely on chemical composition (e.g., Model 2A, Bosque and Sales model) by approximately twofold.

Conclusions:

• The developed empirical models offer efficient prediction of molecular polarizability.
• Model 2E, utilizing an additive hypothesis with 14 atom types, provides superior accuracy compared to other tested methods.
• Both Model 1A and Model 2E are expected to be valuable tools for QSAR and QSPR research.