Predicting the Mutagenic Activity of Nitroaromatics Using Conceptual Density Functional Theory Descriptors and...

Area of Science:

• Computational chemistry
• Toxicology
• Chemical safety

Background:

• Nitroaromatic compounds (NAs) are industrially relevant but pose genotoxic risks.
• Accurate mutagenicity prediction is vital for chemical safety assessments.
• Existing methods require improvement for reliable risk evaluation.

Purpose of the Study:

• To develop and validate a predictive model for nitroaromatic compound mutagenicity.
• To integrate conceptual density functional theory (CDFT) descriptors with machine learning (ML) models.
• To assess the influence of aqueous-phase properties on mutagenicity prediction.

Main Methods:

• Utilized OECD QSAR guidelines for feature selection and model development.
• Employed decision-tree-based algorithms (Random Tree, JCHAID*, SPAARC) and multilayer perceptrons (MLPs).
• Integrated CDFT descriptors and evaluated a novel electronic analog to LogP (LogQP).

Main Results:

• Achieved high predictive accuracy: internal >80% and external ~90%.
• Models demonstrated strong interpretability and robustness.
• Aqueous-phase electronic properties and electrophilicity descriptors outperformed vacuum-based methods.

Conclusions:

• The combined CDFT descriptors and shallow ML models offer a robust, interpretable framework for predictive toxicology.
• This approach enhances chemical risk assessment and supports regulatory applications.
• The study highlights the importance of aqueous-phase properties in predicting NA mutagenicity.