Unraveling the ecotoxicity of micro(nano)plastics loaded with environmental pollutants using ensemble machine...

Micro(nano)plastics are ubiquitous and pose a severe threat to the environment and human health. Despite increasing research, most existing studies have focused on the toxicity of micro(nano)plastics as individual pollutants. Furthermore, fragmented knowledge obtained from separate studies may introduce cognitive biases. Inspired by this, we developed an ensemble machine learning algorithm to predict the combined toxicity (in terms of survival rate) of micro(nano)plastics and environmental pollutants across multiple species. Based on our findings, the following conclusions were drawn: (1) The ensemble machine learning model accurately unraveled the quantitative property-toxicity relationships, achieving strong performance in both cross validation and external validation, with R² > 0.84. (2) Interpretations from the ensemble machine learning model indicated that the combined toxicity is primarily influenced by factors such as pollutant concentration, species, pollutant type, and plastic diameter. (3) Molecular dynamics (MD) simulations further revealed that micro(nano)plastics, after adsorbing the pollutant BDE-47 (2,2',4,4'-tetrabromodiphenyl ether), interact with the cell membranes through van der Waals interactions (less than -200 kJ/mol^-1), ultimately leading to increased membrane damage and deformation. Our modeling results provide an additional avenue for understanding the environmental and health risks associated with micro(nano)plastics and other pollutants, complementing and extending insights gained via experimental methods.