Microbial degradation potential of microplastics in urban river sediments: Assessing and predicting the enrichment of PE/PP-degrading bacteria using SourceTracker and machine learning.

Microplastic mitigation strategies that adapt to various actual aquatic environments require the ability to predict their microbial degradation potential. However, the sources and enrichment characteristics of the degrading bacteria in the plastisphere from river sediments, and their relationship with environmental conditions remain poorly understood. Here, SourceTracker analysis was adopted to investigate the sources and distribution characteristics of total PE/PP-degrading bacteria (TD) and local PE/PP-degrading bacteria (LD) in the plastisphere and surrounding sediments of the urban river. To better characterize the enrichment property of PE/PP-degrading bacteria in the plastisphere, two specific indices, the enrichment ratios of TD (ER) and LD (ER) separately, were first defined in this study. Furthermore, machine learning models were constructed to predict these enrichment ratios. The results showed that river sediments represented an important reservoir of PE/PP-degrading bacteria within the plastisphere (representing 81.8 %). Both the enrichment ratio of TD (R = 0.720) and the enrichment ratio of LD (R = 0.537) showed a significant positive correlation with the carbonyl index of PE/PP, indicating that the enrichment ratios can effectively reflect the microbial degradation potential of PE/PP in sediments. Compared to gradient boosting regression tree, multilayer perceptron, and support vector machines, the random forest (RF) model demonstrated superior accuracy in predicting both the enrichment ratio of TD (R = 0.954, MSE = 0.180) and the enrichment ratio of LD (R = 0.924, MSE = 0.009. It was also observed that the enrichment ratios were higher in river bends, indicating that river bends were potential hot zones for microbial degradation of PE/PP. SHAP analysis highlighted that the key environmental factors exhibited synergistic effects on both enrichment ratios of TD and LD. Finally, the concentration range of key environmental factors that maximize the enrichment ratio was determined. This study constitutes a powerful example of predicting microplastic microbial degradation potential across various scientific disciplines and provides a basis for the effective management of microplastics.