A Comprehensive Mass Spectrometry-Based Workflow for Clinical Metabolomics Cohort Studies

Area of Science:

• Metabolomics
• Clinical Chemistry
• Biomarker Discovery

Background:

• Metabolomics offers comprehensive analysis of biological systems for disease prediction, diagnosis, and prognosis.
• Widespread clinical implementation is hindered by challenges including data complexity, metabolite identification, and reproducibility.

Purpose of the Study:

• To establish a comprehensive and standardized workflow for clinical metabolomics.
• To address key challenges in sample handling, data acquisition, processing, and analysis.

Main Methods:

• Standardized sample collection and preparation across multiple clinical sites.
• Quality control (QC) samples and multiple mass spectrometry (MS) platforms (GC-MS, LC-MS polar, LC-MS lipid) for data acquisition.
• Compound identification using commercial software and in-house libraries (PAppLineTM, UlibMS), batch effect removal (NormAE), and biomarker identification with tree-based models (random forest, AdaBoost, XGBoost).

Main Results:

• Strict adherence to standardized operation procedures (SOP) ensured sample quality.
• Integration of QC samples and advanced analytical techniques improved MS performance and metabolite identification.
• Deep learning and tree-based modeling effectively removed batch effects and identified potential biomarkers, validated by F1 score and a case study.

Conclusions:

• The developed workflow provides a reliable and reproducible method for clinical metabolomics.
• This standardized approach facilitates accurate disease prediction, diagnosis, and prognosis.
• The workflow's validation through a case study demonstrates its practical utility in clinical settings.