Optimized LC-HRMS workflow improves urinary metabolomics reproducibility for bladder cancer profiling
Investigators developed a two-stage design of experiments framework to optimize sample preparation for liquid chromatography-high-resolution mass spectrometry (LC-HRMS) of urine in bladder cancer biomarker discovery. The refined workflow processed 107 clinical samples, yielding 15,344 metabolic signals and 854 putatively identified compounds, with tight quality control clustering confirming strong analytical reproducibility. Comparative analysis of a 50-patient sub-cohort revealed distinct metabolic clustering and significant perturbations in tryptophan metabolism, lipid remodeling, and proteolytic activity associated with disease progression. The study demonstrates that systematic method optimization can enhance the reliability of non-invasive urinary metabolomics, supporting more robust laboratory workflows for future diagnostic panel development.
The original study
A systematic DoE approach for optimizing urinary LC-HRMS metabolomics: enhancing reliability in bladder cancer profiling.
- Authors
- Frolova A, Vokuev M, Ikhalainen Y, Prosuntsova D, Rodin I
- Journal
- Analytical and bioanalytical chemistry
- Type
- Journal Article
- PMID
- 42496706
Original abstract
Untargeted urinary metabolomics presents significant challenges in analytical reproducibility and biological interpretation, particularly in the context of clinical oncology. This study presents a systematically optimized liquid chromatography-high-resolution mass spectrometry (LC-HRMS) workflow for bladder cancer (BCa) biomarker discovery. To address variability in sample preparation, a two-stage design of experiments (DoE) approach was applied to systematically optimize key parameters affecting metabolite extraction efficiency, thereby improving the reproducibility of subsequent non-invasive profiling. The performance of the workflow was evaluated through the systematic assessment of instrumental stability and injection precision using pooled quality control (QC) samples. Following peak picking and alignment, a comprehensive raw dataset of 15,344 metabolic signals was generated, leading to the putative identification of 854 compounds. Unsupervised principal component analysis (PCA) demonstrated reproducible instrumental performance, indicated by tight QC sample clustering. From the total clinical cohort of 107 patients, a demographically matched sub-cohort of 50 individuals was evaluated to suppress confounding physiological noise. This comparative model revealed distinct disease-specific clustering and demonstrated significant perturbations in the tryptophan metabolic axis, membrane lipid remodeling, and enhanced proteolytic activity, characterized by an evident peptide overflow, associated with BCa progression. This systematically optimized methodology provides a reliable analytical approach for identifying non-invasive diagnostic panels, supporting the implementation of efficient laboratory workflows aligned with Analytics 5.0 principles.