Discrepancies in Biomarker Identification in Peak Picking Strategies in Untargeted Metabolomics Analyses of Cells, Tissues, and Biofluids

Different software and algorithms are available for peak picking in nontargeted metabolomics, and each may have its own strengths and limitations. The choice of the peak picking method can significantly influence the results obtained, including the number and identity of metabolites detected, their quantification, and subsequent biomarker analysis. The impact of peak picking by different tools in an untargeted metabolomics-based biomarker study is largely understated. This study compares two popular open-source software tools for peak picking in untargeted metabolomics of cancer cells, tissues, and biofluids: XCMS and MZmine 2. The investigation evaluates the impact of these peak picking algorithms on biomarker identification after careful noise filtering by blank feature filtering (BFF). We found significant discrepancy between the results obtained from XCMS and MZmine 2, regardless of the sample types, solvent gradient phases, retention time, or mass-to-charge ratio (<i>m</i>/<i>z</i>) tolerances used. Notably, this study revealed significant disagreement between peak picking tools in the context of metabolite-based biomarker study after BFF and highlighted the importance of carefully evaluating and selecting appropriate peak picking tools to ensure reliable and accurate results in untargeted metabolomics research.