Developing predictive models for assessing LC₅₀ of organic contaminants in Gammarus species using interpretable...

Area of Science:

• Environmental Toxicology
• Ecotoxicology
• Quantitative Structure-Activity Relationships (QSAR)

Background:

• Gammarus species are vital bioindicators for assessing aquatic ecosystem health and water quality.
• These invertebrates exhibit high sensitivity to various chemical pollutants, making them crucial for toxicity studies.
• Existing quantitative structure-activity relationship (QSAR) models for Gammarus toxicity often rely on complex descriptors.

Purpose of the Study:

• To develop and present four novel predictive models for estimating the median lethal concentration (LC50) of organic compounds in four key Gammarus species.
• To utilize interpretable structural parameters of organic molecules for toxicity prediction, moving beyond computationally intensive descriptors.
• To leverage the largest available experimental dataset for Gammarus toxicity to enhance model accuracy and robustness.

Main Methods:

• Development of four predictive models for Gammarus lacustris, Gammarus fasciatus, Gammarus pulex, and Gammarus pseudolimnaeus.
• Utilized interpretable molecular features such as functional groups, atom types, and structural characteristics.
• Validated models against extensive internal and external experimental toxicity datasets.

Main Results:

• The new models achieved significantly higher R² values compared to existing QSAR models across all tested Gammarus species.
• R² values for the new models ranged from 0.915 to 0.976, demonstrating superior reliability.
• The models effectively captured the relationship between chemical structure and toxicity (pLC50) for these sensitive aquatic invertebrates.

Conclusions:

• The developed predictive models offer a more reliable and robust approach to estimating chemical toxicity in Gammarus species.
• These models provide straightforward tools for water quality assessment and ecological risk evaluation.
• The use of interpretable structural parameters facilitates a better understanding of the mechanisms driving chemical toxicity in aquatic invertebrates.