Making waves: From molecular inventories to mechanistic networks - an isotope-anchored reactomics framework for...

Current disinfection byproduct (DBP) research, despite generating vast molecular inventories via ultrahigh-resolution mass spectrometry, remains constrained by a "data-rich, insight-poor" paradigm dominated by static compositional analysis. Exhaustively cataloging individual DBPs within these complex matrices is an endless endeavor; the critical bottleneck lies in resolving the dynamic reaction networks that drive their formation from natural organic matter (NOM). This study proposes a transformative framework that synergistically integrates isotope labeling, paired mass distance (PMD)-based reaction network analysis, and explainable artificial intelligence (XAI) to navigate the NOM-DBP continuum. Within this triad, stable isotope labeling acts as a molecular tracking system, providing direct experimental evidence for atom incorporation and anchoring the identification of precursor-product relationships. These validated links could serve as the foundation for constructing confidence-ranked PMD reaction networks, revealing the interconnected transformation pathways and prioritizing highly connected hub precursor features for further validation. XAI is then applied to decode the chemical logic of these networks, translating statistical correlations into actionable, mechanism-based hypotheses. As a proof of concept, we further demonstrate that isotope-anchored PMD features can enrich the candidate set with reaction-relevant byproduct formulas, with XAI used to interpret their contributions. This integrated approach offers a pathway for shifting DBP research from reactive, end-of-pipe compliance monitoring toward more predictive and proactive intervention strategies. We envision this framework as the cornerstone for next-generation precision water engineering and digital twin systems, enabling the development of source-targeted DBP control through iterative experimental validation and model refinement.