Integrating Experimental Toxicology and Machine Learning to Model Levonorgestrel-Induced Oxidative Damage in...

Area of Science:

• Environmental Toxicology
• Biomarker Discovery
• Computational Toxicology

Background:

• Levonorgestrel (LNG) is an emerging aquatic contaminant.
• LNG can disrupt redox homeostasis and induce oxidative stress in aquatic organisms.
• Understanding LNG's sublethal effects requires sensitive detection methods.

Purpose of the Study:

• To investigate tissue-specific and dose-time dependent oxidative stress responses in zebrafish exposed to LNG.
• To identify reliable biomarkers for LNG-induced oxidative damage.
• To apply machine learning for predictive modeling of contaminant effects.

Main Methods:

• Zebrafish were exposed to environmentally relevant LNG concentrations (0.312 µg/L and 6.24 µg/L) for 24, 48, and 96 hours.
• Redox biomarkers (SOD, CAT, GPx, MDA) were quantified in liver and muscle tissues.
• Supervised machine learning algorithms (including Gradient-Boosted Trees and XGBoost) were used for data analysis and prediction.

Main Results:

• LNG exposure, particularly at higher concentrations, induced significant changes in oxidative stress biomarkers.
• Hepatic glutathione peroxidase (GPx) activity increased markedly and was identified as a sensitive indicator of LNG exposure.
• Machine learning models accurately predicted exposure outcomes, with hepatic GPx being the most informative feature.

Conclusions:

• Hepatic GPx serves as a sentinel biomarker for detecting levonorgestrel-induced oxidative stress in aquatic ecosystems.
• Machine learning approaches enhance toxicological frameworks for predicting sublethal contaminant effects with high temporal resolution.
• This study highlights the potential of integrated toxicogenomics and machine learning in environmental risk assessment.