Machine learning in ecotoxicology: Pollutant exposure levels and detection, biotoxicity and environmental behavior...

Area of Science:

• Ecotoxicology
• Environmental Science
• Computational Toxicology

Background:

• Environmental pollution necessitates advanced methods beyond traditional toxicological assays.
• Machine learning (ML) offers powerful tools for analyzing complex environmental data.
• Limitations in current ecotoxicology methods drive the adoption of data-driven approaches.

Purpose of the Study:

• To highlight the accelerating adoption and advantages of ML in ecotoxicology.
• To detail ML applications in predicting pollutant exposure and toxicity.
• To discuss ML's role in modeling pollutant behavior and enhancing ecological validity.

Main Methods:

• Utilizing large-scale heterogeneous datasets for ML model construction.
• Applying ML for toxicity prediction across single-, multi-, and cross-species.
• Employing interpretability methods (SHAP, LIME) for mechanistic insights.
• Integrating species traits and environmental covariates for refined predictions.

Main Results:

• ML enables rapid and accurate prediction of pollutant exposure and toxicity.
• Data-driven models enhance toxicity prediction and support extrapolation to new species.
• ML models pollutant environmental behavior under dynamic conditions.
• Standardized preprocessing and interpretability methods improve model reproducibility and understanding.

Conclusions:

• ML significantly advances ecotoxicology by improving prediction accuracy and ecological relevance.
• Standardization, interpretability, and integration of environmental data are key for robust ML applications.
• ML provides a powerful, reproducible, and regulatory-relevant framework for environmental risk assessment.