来源：ACS Publications
Rare earth elements (REEs) are emerging contaminants with escalating environmental releases. However, REE health risk assessment faces critical challenges due to inconsistent cytotoxicity benchmarks and complex multielement exposures. Here, we developed a comprehensive framework to systematically assess the cytotoxic risks of 16 rare earth ions (REIs). The framework integrates single and binary exposures of 16 REIs across eight human cell lines, machine learning models, and population exposure scenario predictions. The high-throughput screening of single exposure on three end points revealed common disruption of cellular energy metabolism, identifying ATP depletion as a robust benchmark point of departure (BPOD). On the basis of the BPOD, 120 binary combinations of 16 REIs indicated predominantly antagonistic interactions among REEs. The consensus machine learning model trained on the single and binary exposure data sets demonstrated robust predictive performance for mixture cytotoxicity. Application of the model to human exposure scenarios showed high risks for the population in mining areas and negligible risks for the general population. The significant linear correlation of cytotoxic responses between cell lines and human primary hepatocytes confirmed the human health relevance and model reliability. Our study provides a toolkit to establish a standardized, scalable, and integrative risk assessment for REEs and other emerging multielement contaminants.