The Monte Carlo technique as a tool to predict LOAEL

Area of Science:

• Toxicology
• Computational Chemistry
• Medicinal Chemistry

Background:

• Quantitative Structure-Activity Relationships (QSARs) are crucial for predicting chemical compound toxicity.
• Accurate prediction of the Lowest Observed Adverse Effect Level (LOAEL) is essential for risk assessment.
• Molecular structure representation using Simplified Molecular Input-Line Entry Systems (SMILES) is widely adopted.

Purpose of the Study:

• To develop QSAR models for predicting the LOAEL of organic compounds.
• To introduce a novel criterion for evaluating QSAR model estimation quality.
• To establish a methodology for defining the domain of applicability in QSAR modeling.

Main Methods:

• Utilized a dataset of 341 organic compounds represented by SMILES notation.
• Developed QSAR models using training and external validation sets.
• Correlated estimation quality criteria with predictive potential (root-mean-square error).
• Defined mechanistic interpretation and domain of applicability based on a probabilistic viewpoint.

Main Results:

• Developed statistically sound one-variable QSAR models for LOAEL prediction.
• Identified a correlation between the proposed criterion and QSAR model predictive performance.
• Successfully defined the domain of applicability for the developed QSAR models.
• Presented a methodology for applicability domain definition using SMILES-based optimal descriptors.

Conclusions:

• The proposed criterion effectively assesses QSAR model quality and predictive potential.
• QSAR models developed using SMILES notation can accurately predict LOAEL values.
• The methodology provides a robust framework for defining the domain of applicability in QSAR studies.
• This approach enhances the reliability and interpretability of QSAR models in toxicology.