Authors
Yizhe Ding, Qiya Sun, Nuo Peng, Qian Ping, Xi Wang, Hao Chen, Lin Wang, Yongmei Li
Published in
Water research. Volume 309. Issue Pt A. Pages 127093. Oct 06, 2026. Epub Oct 06, 2026.
Abstract
In this study, a machine learning-based framework was developed to identify risk features associated with the transformation of extracellular polymeric substances (EPS) during ultraviolet/peracetic acid (UV/PAA) disinfection, which had been previously underestimated. Molecular‑level analysis revealed that EPS transformation follows a time‑dependent pathway: aromatic and highly unsaturated species undergo preferential cleavage initially, followed by oxidation, heteroatom redistribution, and molecular rearrangement, ultimately leading to slower fragmentation within a residual oxygen‑rich matrix. Specifically, whereas fission reactions initially dominated protein-to-lipid (63.0%) and protein-to-lignin (38.7%) transformations, lipid-derived oxygenation (53.1%-59.4%) and lignin-associated fission (25.0%-30.9%) contributed consistently throughout the reaction. Notably, although UV/PAA disinfection reduced the overall risk associated with EPS transformation, safety risks still persisted. Machine learning analysis identified the O/C ratio, octanol-water partition coefficient, and topological polar surface area as the dominant molecular features governing risk assessment, with mean SHapley Additive exPlanations (SHAP) values of 0.69, 0.62, and 0.40, respectively. These features mechanistically link hazard evolution to oxygenation, hydrophobicity, and molecular polarity. Collectively, these findings indicate that research on EPS transformation during UV/PAA disinfection should move beyond a mechanism-focused paradigm toward a hazard-based risk assessment framework. Accordingly, this study not only provides a data-driven tool for assessing the risks associated with EPS transformation but also offers a practical framework for comprehensively evaluating its potential as a safer disinfection technology.
PMID:
42860086
Bibliographic data and abstract were imported from PubMed on 12 Oct 2026.
Read full publication at:
Please sign in
to see all details.
Advertisement
Stats
- Recommendations n/a n/a positive of 0 vote(s)
- Views 12
- Comments 0