Civil Engineering Departmental Seminar - Advancing Analysis of Halogenated and Nitrogenous Contaminants in Water

Advancing Analysis of Halogenated and Nitrogenous Contaminants in Water

Water disinfection effectively prevents microbial infections, but it inevitably produces disinfection byproducts (DBPs). To date, halogenated DBPs largely remain unidentified, although 50 years of research has significantly improved water quality. Nontargeted high-resolution mass spectrometry (HRMS) analysis of water detects large datasets of chemical features, but structural annotation is a bottleneck in advancing DBP identification. To address this challenge, we have developed two machine learning–guided platforms, HalogenFinder and IodoFinder, to automatically recognize Cl-, Br-, and I-containing compounds in LC-HRMS nontargeted analysis datasets. HalogenFinder employs hierarchical Random Forest models trained on simulated isotope patterns from 1.3 million PubChem molecular formulas, achieving 95% accuracy in identification of Cl/Br-compounds and number of halogen atoms in a chemical feature. IodoFinder leverages diagnostic fragmentation features from 6,000 MS/MS spectra of 381 iodinated compounds, reaching 96.5% and 100% accuracy in positive and negative ionization modes, respectively. HalogenFinder and IodoFinder were validated using authentic standards (277 Cl/Br-compounds; 161 I-compounds) and applied to real water samples and DBP mixtures. These tools revealed novel halogenated DBPs. Additionally, isotope labeling combined with HDPairFinder enabled identification of reactive nitrogenous precursors in source water. Together, these platforms offer high-throughput, accurate, and accessible solutions for comprehensively identifying halogenated contaminants and reactive precursors in water, enabling exposomic studies of water contaminants.