Metabolomics workbench: data-driven classification and correlation analysis between metabolite classes, biomarkers, and drugs.

Correlation analysis of 6752 blood metabolites reveals that structurally related clusters have up to 0.6 correlation in abundance across human studies.

• Why it matters: Large-scale metabolomics repositories are underutilized for systematic re-analysis, limiting the discovery of new biological insights from existing data.
• What they did: Analyzed 44,165 samples from human blood studies using Orbitrap or QTOF mass spectrometry, harmonized metabolites with PubChem, embedded them with SMILES, and performed correlation analysis on 337,120 metabolite pairs.
• The result: Findings include correlated metabolite clusters, candidate biomarkers with correlations >0.7, and metabolic signatures for drugs like acetaminophen, demonstrating the value of large-scale mining and software tools for biomedical discovery.