Modeling ready biodegradability of fragrance materials

Area of Science:

• Environmental Chemistry
• Computational Chemistry
• Toxicology

Background:

• Assessing the environmental impact of chemicals is crucial.
• Biodegradability prediction is essential for fragrance materials.
• Existing models may lack specificity for complex chemical mixtures.

Purpose of the Study:

• Develop accurate quantitative structure-activity relationship (QSAR) models for predicting ready biodegradability.
• Evaluate classification and regression tree (CART) and k-nearest neighbors (kNN) for this purpose.
• Compare model performance against established methods like BIOWIN.

Main Methods:

• Quantitative structure-activity relationships (QSAR) were employed.
• Two classification algorithms, CART and kNN, were utilized for modeling.
• Models were validated using external prediction sets and the leverage approach for applicability domain verification.

Main Results:

• High internal and external validation metrics were achieved: sensitivity (≥68%), specificity (≥73%), and overall accuracy (≥75%).
• The developed QSAR models demonstrated superior predictive performance compared to BIOWIN6, especially for non-readily biodegradable fragrance materials.
• The structural applicability domain was successfully verified.

Conclusions:

• Specific QSAR models for fragrance materials offer improved biodegradability prediction accuracy.
• CART and kNN are effective methods for developing such predictive models.
• These findings contribute to a better environmental risk assessment of fragrance ingredients.