QSAR as a random event: a case of NOAEL

Area of Science:

• Computational Chemistry
• Toxicology
• cheminformatics

Background:

• Predicting chemical toxicity is crucial for risk assessment.
• Quantitative Structure-Activity Relationships (QSAR) offer a computational approach to estimate toxicological endpoints.
• No Observed Adverse Effect Levels (NOAELs) are key metrics in toxicology.

Purpose of the Study:

• To develop and validate QSAR models for predicting NOAELs.
• To establish predictive models using Simplified Molecular Input Line Entry System (SMILES) for molecular representation.
• To assess model performance across multiple random data splits according to OECD principles.

Main Methods:

• Development of one-variable QSAR models using the Monte Carlo method.
• Utilized Simplified Molecular Input Line Entry System (SMILES) for molecular structure input.
• Employed three distinct random splits of the dataset into training, calibration, and validation sets.

Main Results:

• Achieved good statistical performance across all three splits, with QSAR models demonstrating robust predictive power.
• Training set R-squared values ranged from 0.679 to 0.718, and Q-squared values ranged from 0.672 to 0.712.
• Validation set R-squared values were between 0.54 and 0.669, indicating reliable external prediction capabilities.

Conclusions:

• The developed QSAR models provide a reliable method for predicting NOAELs.
• The models adhere to OECD principles, ensuring their applicability in regulatory contexts.
• This approach facilitates efficient in silico assessment of chemical safety.