Integration of QSAR models for bioconcentration suitable for REACH

Area of Science:

• Environmental Chemistry
• Computational Toxicology
• Regulatory Science

Background:

• Quantitative Structure Activity Relationship (QSAR) models offer alternatives to animal testing for REACH regulations.
• Bioconcentration factor (BCF) prediction is a key endpoint where QSAR models show promise.
• Existing BCF models like CAESAR and Meylan have limitations in prediction accuracy.

Purpose of the Study:

• To develop an integrated QSAR model for more reliable BCF predictions.
• To enhance prediction accuracy by combining multiple models and an Applicability Domain Index (ADI).
• To reduce the need for animal testing in chemical safety assessments.

Main Methods:

• Integration of CAESAR and Meylan BCF models.
• Utilization of the VEGA software's Applicability Domain Index (ADI).
• Application of rule-based selection for reliable and conservative predictions, including outlier identification.

Main Results:

• The integrated model achieved R(2)=0.80, RMSE=0.61 log units, and 76% sensitivity for 851 compounds.
• This represents a significant improvement over individual CAESAR (R(2)=0.63) and Meylan (R(2)=0.66) models.
• Predictions within the integrated ADI showed R(2)=0.92 and 85% sensitivity, with RMSE=0.44 log units.
• Applying safety thresholds to "suspicious" compounds yielded 100% sensitivity.

Conclusions:

• Integrated QSAR models, combined with ADI and safety thresholds, provide robust BCF predictions.
• This approach enhances reliability and reduces uncertainty in chemical risk assessment.
• The developed integrated model is a valuable tool for REACH compliance and reducing animal testing.