MESSAR

MESSAR recommends molecular substructures from tandem mass spectra to generate structural hypotheses for unknown metabolites in non-targeted metabolomics.

Key Features:

• Automated substructure recommendation: Uses association rule mining to identify relationships between spectral features in tandem mass spectra and molecular substructures.
• Data-driven learning from public spectral libraries: Learns associations between spectral data and known substructures from public spectral libraries and applies them to unknown spectra.
• Database-independent hypothesis generation: Operates without requiring predefined candidate metabolites or reference spectra, enabling partial identification of "unknown unknowns".

Scientific Applications:

• Hypothesis generation: Enables generation of structural hypotheses for unknown metabolites detected in non-targeted metabolomics studies.
• Partial identification of complex mixtures: Assists in proposing plausible substructures within complex mixtures for further analytical follow-up.
• Application to rare or novel compounds: Supports analysis of metabolites not present in existing reference libraries by recommending substructures from learned associations.

Methodology:

Association rule mining is applied to tandem mass spectra in public spectral libraries to learn associations between spectral features and known substructures, and the learned rules are used to recommend plausible substructures for unknown spectra.