Molecular-level transformation of algal organic matter during water treatment processes by FT-ICR MS integrated with reactomics and interpretable machine learning.

Algal organic matter (AOM) released during cyanobacterial blooms can significantly challenge drinking-water treatment, including an elevated risk of disinfection byproduct (DBP) formation, yet the molecular-scale fate of bloom-derived AOM through treatment remains poorly constrained. Here, we integrated ultrahigh-resolution FT-ICR MS with paired-mass-difference (PMD) reactomics and interpretable machine learning (IML) to resolve (i) bloom-driven molecular transformations, (ii) selective removal by coagulation (COA) and granular activated carbon (GAC), (iii) formula-level chlorination reactivity and DBP formation. Laboratory bloom simulations using Microcystis aeruginosa resulted in the detection of 3373 bloom-derived formulas enriched in nitrogen and reduced character (higher N/C and H/C; lower O/C and aromaticity). Reactomics networks indicated dominant putative transformations involving CHO moieties (e.g., CH2, CO, CH2O) and amine-related changes, with prominent amino-acid-like mass differences. COA (polyferric sulfate) and GAC both substantially reduced bulk dissolved organic carbon (DOC) but generated distinct residual molecular spaces: COA treatment left lower-molecular-weight (MW) and higher-aromatic index (AImod) residues than that of GAC. Supervised ML models identified MW and heteroatom ratios (N/C, S/C, O/C) as key predictors of operationally defined formula-level reactivity following chlorination. Correspondingly, chlorination generated matrix- and treatment-dependent chlorinated organic compounds (COCs), including nitrogen-containing chlorinated features uniquely detected in AOM-containing waters. These findings demonstrate that conventional treatment processes, while effective at reducing bulk organic matter, leave distinct residual precursor pools that can alter COCs formation under bloom conditions. This molecular-level framework provides new insights into precursor-level compositional changes and offers a basis for evaluating DBP formation and precursor control strategies in cyanobacterial bloom-impacted drinking-water system.