Learning from All Views: A Multiview Contrastive Framework for Metabolite Annotation.

Area of Science:

• Metabolomics
• Computational Chemistry
• Bioinformatics

Background:

• Metabolomics, utilizing high-throughput mass spectrometry, is crucial for understanding cellular biochemistry and disease mechanisms.
• Low molecular structure assignment rates due to spectral complexity hinder advancements in metabolomics.
• Current methods struggle to effectively integrate diverse data views for spectral annotation.

Purpose of the Study:

• To introduce MultiView Projection (MVP), a novel framework for spectral annotation in metabolomics.
• To develop a method that learns a joint embedding space between molecules and spectra.
• To improve the accuracy and robustness of molecular structure identification from mass spectrometry data.

Main Methods:

• MVP employs contrastive multiview learning, integrating molecular graphs, fingerprints, spectra, and consensus spectra.
• The framework learns a joint embedding space by capturing mutual information across multiple data views.
• MVP supports flexible annotation using individual or consensus spectra.

Main Results:

• MVP significantly improves molecular candidate ranking by learning from all data views jointly.
• Annotation using consensus spectra with MVP outperforms rank aggregation strategies.
• MVP achieved 35.99% rank@1 for mass-based retrieval and 13.96% for formula-based retrieval using consensus spectra.

Conclusions:

• MVP provides a flexible and extensible foundation for spectral annotation in metabolomics.
• The framework enhances the potential of mass spectrometry for disease mechanism and drug development studies.
• Improved spectral annotation rates facilitated by MVP can accelerate personalized medicine initiatives.