Exploiting the gut microbiota of aquatic animals as indicators of microplastic pollution using interpretable machine learning models

The response of aquatic animal gut microbiota to microplastics has been extensively studied and shows sensitivity, however, the potential of using gut microbiota as indicators for microplastic pollution has not yet been fully explored. To address this gap, we analyzed publicly available sequencing data of gut microbiota from 17 aquatic species (634 samples), including both microplastic-exposed and unexposed groups. Using interpretable machine learning models, we demonstrated that microplastics significantly altered the composition and functional profiles of gut microbiota in both freshwater and saltwater animals, reducing functional redundancy. Random forest models based on genus-level taxonomy effectively indicated microplastic pollution, achieving AUC (Area Under the receiver-operating-characteristic Curve) values above 0.8. Through SHAP (SHapley Additive exPlanation) value analysis, we identified 20 freshwater and 25 saltwater gut microbial indicators of microplastic pollution, with Aurantimicrobium and Salipiger_489036 emerging as the most important indicators for freshwater and saltwater environments, respectively. Notably, the importance of these indicators varied depending on the host species and pollution context. For example, Aurantimicrobium exhibited greater importance under conditions of prolonged exposure and higher microplastic concentrations, whereas Salipiger_489036 was sensitive to microplastic size. Overall, our findings highlight the potential of gut microbiota as a novel tool for assessing microplastic pollution in aquatic environments.