We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Distribution and Property Prediction of Microplastics in Landfill Leachate under Different Treatment Processes Using Knowledge Graph-Enhanced Molecular Contrastive Learning
Summary
Researchers tracked microplastic distribution and transformation across treatment stages of a full-scale landfill leachate system (membrane bioreactor, ultrafiltration, nanofiltration, reverse osmosis). Microplastic concentrations fell from 14,150 items/L in raw leachate to 4,400 items/L in concentrate, and a knowledge graph-enhanced machine learning model predicted toxicity properties of microplastics at each treatment stage.
Landfill leachate serves as a critical sink for microplastics (MPs), whose persistence, heterogeneity, and potential toxicity pose growing environmental and human health concerns. However, the fate and transformation of MPs across different leachate treatment processes remain poorly characterized, and traditional toxicity assessments are labor-intensive, time-consuming, and costly. In this study, we investigated the distribution, physicochemical properties, and toxicological behaviors of MPs throughout a full-scale leachate treatment system, encompassing membrane bioreactors, ultrafiltration, nanofiltration, reverse osmosis, and the double tube reverse osmosis (DTRO) concentrates. Results revealed the total MPs concentration decreased from 14,150 items/L in the influent to 4400 items/L in the RO effluent, with an overall removal efficiency of 68.91 %. The total MPs concentration decreased from 14,150 items/L in the influent to 4400 items/L in the RO effluent, with an overall removal efficiency of 68.91 %. Among the 14 detected MPs types, cellophane and polyamide were predominant, accounting for 74.61 % and 23.71 % of total MPs, respectively. Using a novel unsupervised model-KANO (knowledge graph-enhanced molecular contrastive learning)-we predicted toxicological responses of 14 representative MPs. High toxicity potentials were observed for gastrointestinal (up to 95.62 %), immune system (up to 84.66 %), and skin-related disorders (up to 81.34 %), particularly in MBR influent and DTRO concentrates. This study provides a detailed characterization of their distribution and predicted toxicity across engineered treatment processes.