Machine learning uncovers tidal DOM transformations and keystone molecules via FT-ICR MS and reactomics for estuarine...

Tidal cycles in estuaries dynamically regulate the composition and transformation of dissolved organic matter (DOM). However, conventional methods exhibit inadequate capacity to decipher the molecular transformation pathways, thereby limiting the understanding of nitrogen-sulfur biogeochemical cycles therein. This study employed Fourier transform ion cyclotron resonance mass spectrometry (FT-ICR MS), reaction omics based on paired mass distance (PMD) networks, and machine learning (ML) approaches to investigate tidal-driven DOM dynamics in the estuarine sediments of a representative mountainous river, i.e., Mulan River. Results revealed that tidal cycling significantly enhanced the humification of DOM in sediments, with O/C ratio increased from 0.251 for shallow layer samples when tide receded to 0.395 in deep layer sample collected at high tide. Tide cycles also promoted the accumulation of nitrogen-containing and sulfur-containing compounds, i.e., CHON and CHONS moieties, particularly in deeper sediments (proportion up to 37 %). ML models, i.e., XGBoost and LightGBM identified high molecular weight (> 450 Da), elevated N/C (> 0.05), and S/C (> 0.025) ratios as key predictors of biodegradable DOM. PMD-based reaction networks uncovered microbially mediated transformations, including dealkylation, amide hydrolysis, and desulfonation, driving dominant fractions shifting from aliphatic/proteins to lignin/carboxyl-rich alicyclic molecules during tidal events. Network topology analysis disclosed that CHOS compounds, e.g., C₁₇H₂₀O₈S₁, emerged as pivotal nodal regulators of sulfur cycling, serving as metabolic hubs bridging aerobic and anaerobic microbial communities. Putatively derived from cysteine/methionine biotransformation products, these sulfur-enriched molecules exhibited significantly enhanced betweenness centrality in post-tidal reaction networks, underscoring their role in maintaining functional resilience under oscillating redox regimes.