Machine Learning in Environmental Toxicology: Predicting Endocrine Disruption and Cross-Species Toxicity of Emerging Contaminants

Introduction:
The rapid proliferation of synthetic chemicals, pharmaceuticals, and agrochemicals has transformed global ecosystems, raising concerns about endocrine-disrupting chemicals (EDCs) and cross-species toxicity. Traditional in vivo toxicological testing is increasingly inadequate to evaluate the growing volume of emerging contaminants (ECs), necessitating computational approaches.

Methods:
This systematic review evaluates the application of artificial intelligence (AI) and machine learning (ML) in environmental and veterinary toxicology within a One Health framework. Databases spanning 2000–2026 were analyzed to assess predictive modeling of EDCs, including PFAS, phthalates, bisphenols, and pharmaceutical mixtures. Classical ensemble algorithms (Random Forest, GPBoost) were compared with deep learning architectures such as Communicative Message Passing Neural Networks (CMPNN) and 3D-structure-based models (3DMol-Tox). Cross-species toxicity prediction was examined by juxtaposing Interspecies Correlation Estimation (ICE) models with multi-feature ML algorithms.

Results:
AI-driven models demonstrated superior accuracy in predicting endocrine disruption and extrapolating toxicity across taxonomic boundaries. Advanced deep learning approaches outperformed traditional methods in handling complex chemical structures and mixture effects. Critical methodological challenges were identified, including information leakage in train-test splitting, which can inflate predictive performance. Integrating AI with ecotoxicological data enables more reliable risk assessment, linking molecular initiating events to sentinel species, veterinary health, and ecosystem resilience.

Conclusions:
Machine learning provides a transformative framework for predictive toxicology, bridging gaps between molecular mechanisms and ecosystem-level outcomes. These approaches enhance cross-species risk assessment, accelerate screening of emerging contaminants, and support One Health strategies for safeguarding environmental and animal health.