Enhanced modeling of chlorine-organic reaction kinetics to assess the fate of environmental chemicals in disinfection process

Document Type

Article

Publication Date

3-15-2026

Abstract

The escalating release of emerging organic contaminants into aquatic systems necessitates rigorous investigation of their reactivity with chlorine, a ubiquitous disinfectant in water treatment. This study establishes the most comprehensive dataset to date on chlorine-organic reaction kinetics, integrating novel high-throughput microplate assays with literature data to compile 354 compounds. Advanced machine learning models—a regression-based extreme gradient boosting and a classification-based random forest, both utilizing modified molecular fingerprints—achieved superior predictive accuracy (validation-set R2 = 0.82, balanced accuracy = 0.903), outperforming prior quantitative structure-activity relationship (QSAR) models by 35–50%. To address frequency-dependent artifacts in SHAP interpretation, a novel reactivity index (RI) was developed, enabling unbiased quantification of substructure-specific contributions to chlorination kinetics. RI captured reactive low-frequency moieties (e.g., sulfur), which were overlooked by SHapley Additive exPlanations (SHAP) due to frequency bias. Application of these models to 10,589 environmentally relevant organic compounds revealed critical reactivity trends: 43% of pesticides exhibited slow reactivity (half-life > 6.8 h), while 22% of endocrine-disrupting chemicals showed fast degradation (half-life < 4.1 min). Mechanistic analysis identified electron-rich functional moieties. Notably, density functional theory calculations demonstrated that -COO- lowers reaction barriers by 11.6 kcal/mol compared to -COOH, challenging conventional electron-withdrawing classifications. Sulfur moieties, overlooked in prior studies, emerged as potent reactivity drivers. These findings provide a framework for predicting contaminant fate during chlorination, with direct implications for optimizing water treatment processes and prioritizing environmental monitoring of persistent pollutants. The integration of explainable machine learning with kinetic modeling advances molecular-level understanding of chlorine-organic interactions, establishing a foundation for next-generation contaminant mitigation strategies.

Publication Title

Water Research

Link to Full Text

Share

COinS