Monte Carlo optimization-based QSPR modeling of molar refractivity: Descriptor stability and applicability domain...

Area of Science:

• Computational Chemistry
• Cheminformatics
• Quantitative Structure-Property Relationships (QSPR)

Background:

• Traditional fragment-constant methods for predicting molecular properties are limited.
• There is a need for data-driven QSPR models with robust statistical validation.

Purpose of the Study:

• To develop an additive fragment-based QSPR model for molar refractivity (MR).
• To integrate Monte Carlo optimization and statistical control for enhanced model performance.
• To define and assess an applicability domain based on descriptor representativeness.

Main Methods:

• Developed an additive QSPR model using fragment-level attributes.
• Employed Monte Carlo optimization within the CORAL framework.
• Utilized multiple independent training-validation splits for statistical control.
• Defined an applicability domain using statistical defect metrics.

Main Results:

• Achieved high coefficients of determination (R² ≈ 0.86–0.89) in both training and external validation sets.
• Demonstrated consistent root-mean-square errors, indicating reproducible model behavior.
• The applicability domain identified statistically under-supported compounds, correlating with minor prediction error increases.

Conclusions:

• Monte Carlo-optimized additive modeling provides a statistically audited extension of fragment-constant QSPR approaches.
• The developed workflow integrates robustness analysis and applicability domain assessment transparently.
• This method enhances the reliability and interpretability of QSPR predictions for molar refractivity.