By integrating previously overlooked drivers AI boosts bioaccumulation assessment in fish

Area of Science:

• Environmental Chemistry
• Toxicology
• Regulatory Science

Background:

• The bioconcentration factor (BCF) is increasingly used in chemical regulation to assess bioaccumulation potential.
• Current regulatory approaches often assume BCF is a fixed, chemical-specific property.

Purpose of the Study:

• To develop a machine learning model for estimating experimental BCFs.
• To investigate the influence of exposure concentration and duration on BCF.
• To propose new metrics for improved bioaccumulation assessment in chemical regulation.

Main Methods:

• A machine learning model was trained using chemical properties and test conditions for 962 chemicals.
• The model estimated BCFs for 16 fish species, achieving up to 90% accuracy.
• A review of 165 regulatory studies submitted to EU authorities was conducted.

Main Results:

• The study demonstrated that BCF is not a fixed criterion but increases with lower exposure concentration and longer exposure duration.
• Approximately half of the chemicals reviewed were not classified as bioaccumulative based on experimental BCFs, despite meeting worst-case criteria.
• The proposed metrics, BCF50 and BCF90, offer objective measures for bioaccumulation screening.

Conclusions:

• The findings challenge the assumption of BCF as a static chemical property.
• Current regulatory practices may misclassify bioaccumulative substances due to the dynamic nature of BCF.
• Implementing BCF50 and BCF90 into regulations can enhance the accuracy and objectivity of bioaccumulation assessments.