MS2Query: reliable and scalable MS² mass spectra-based analogue search

Area of Science:

• Chemistry
• Biochemistry
• Data Science

Background:

• Metabolomics research is limited by the difficulty of identifying and annotating metabolites.
• Existing spectral libraries contain few annotated spectra, leading to limited identification of metabolites.
• Searching for chemical analogues, which are similar but not identical molecules, is a promising alternative for metabolite annotation.

Purpose of the Study:

• To develop and evaluate MS2Query, a novel tool for enhanced metabolite identification in complex mixtures.
• To improve the reliability and speed of searching for metabolite analogues and exact matches in spectral libraries.
• To increase the overall metabolite annotation rate in metabolomics studies.

Main Methods:

• MS2Query utilizes machine learning, integrating mass spectral embedding-based chemical similarity predictors (Spec2Vec and MS2Deepscore).
• The tool incorporates detected precursor masses to rank potential analogues and exact matches.
• Performance was benchmarked using reference mass spectra and experimental case studies.

Main Results:

• MS2Query demonstrates improved reliability and scalability compared to existing methods.
• The tool effectively ranks potential metabolite analogues and exact matches.
• Benchmarking confirmed the enhanced performance on reference and experimental data.

Conclusions:

• MS2Query significantly enhances the annotation rate of metabolomics profiles.
• The tool offers a more reliable and efficient approach to metabolite identification.
• MS2Query facilitates the discovery of new biological insights through improved metabolomics analysis.