Development of classification and regression based QSAR models to predict rodent carcinogenic potency using oral...

Area of Science:

• Toxicology
• Computational Chemistry
• Chemometrics

Background:

• Carcinogenicity is a major human health concern.
• Oral slope factors (OSFs) quantify carcinogenic potency from oral exposure.
• Regulatory agencies use quantitative structure-activity relationship (QSAR) models to address data gaps.

Purpose of the Study:

• Develop and validate QSAR models for rodent carcinogenicity.
• Create a carcinogenicity classification model using Linear Discriminant Analysis (LDA).
• Provide quantitative structure-carcinogenicity interpretations and identify safer chemicals.

Main Methods:

• Developed regression and LDA models for 70 diverse chemicals.
• Assessed models using OECD QSAR validation principles.
• Validated models internally and externally.
• Utilized Pharmacological Distribution Diagrams (PDDs) for visualization.

Main Results:

• Established validated in silico quantitative structure-carcinogenicity regression models.
• Developed a robust LDA model for carcinogenicity classification.
• Achieved quantitative interpretation of structural carcinogenicity information.
• Identified discriminant functions differentiating lower and higher carcinogenicity.

Conclusions:

• The developed QSAR and LDA models provide reliable predictions of rodent carcinogenicity.
• In silico studies offer valuable insights into structural determinants of carcinogenicity.
• PDDs aid in selecting chemicals with reduced carcinogenic potential.
• The models support regulatory risk assessment and chemical safety evaluations.