BOOK OF ABSTRACTS 33 IHSS&IWA26 / BRNO / CZECHIA / 23–28 August 2026 Monday, 24 August 2026 / Hall B+C Analysis and Characterization preserving real chemical differences between sample groups. While Systematic Error Removal Using Random Forest (SERRF) was best at making the QC tightening, meaning it most strongly improved technical consistency across the run. The results also showed that preprocessing, formular-assignment and feature-definition choices had a substantial impact on the usable feature space and therefore on the robustness of the normalization comparison. Overall, this work shows that batch effects are a major issue in NOM characterization and that their treatment requires a balanced correction strategy rather than reliance on QC precision alone. Acknowledgement We acknowledge the Bioanalytical Mass Spectrometry Facility, UNSW that funded this work through a student scholarship.
RkJQdWJsaXNoZXIy NDA4Mjc=