mWISE: An Algorithm for Context-Based Annotation of Liquid Chromatography-Mass Spectrometry Features through...

Area of Science:

• Metabolomics
• Bioinformatics
• Computational Biology

Background:

• Untargeted metabolomics using liquid chromatography coupled to mass spectrometry (LC-MS) enables the detection of numerous metabolites.
• A significant bottleneck in metabolomics is the accurate and confident annotation of detected metabolites.
• Current annotation methods often identify only a small fraction of the metabolites present in biological samples.

Purpose of the Study:

• To introduce mWISE (metabolomics wise inference of speck entities), an R package designed for context-based annotation of LC-MS data.
• To improve the accuracy and efficiency of metabolite identification in untargeted metabolomics studies.
• To provide a robust tool for prioritizing candidate metabolites through graph diffusion algorithms.

Main Methods:

• mWISE utilizes a three-step algorithm: mass-to-charge ratio matching to the KEGG database, clustering and filtering of potential candidates, and graph diffusion for final prioritization.
• Performance evaluation involved three public LC-MS datasets across positive and negative ionization modes.
• Comparison with existing annotation tools, including xMSannotator, assessed performance and computational time.

Main Results:

• mWISE demonstrated superior performance and reduced computation time compared to xMSannotator across tested datasets.
• A diffusion configuration of mWISE achieved a mean sensitivity of 0.63 (0.07), outperforming xMSannotator's 0.55 (0.19).
• mWISE proposed chemical structures closer to the original compounds, indicating higher annotation accuracy.

Conclusions:

• mWISE significantly enhances the metabolite annotation process in untargeted metabolomics.
• The graph diffusion approach is a key component for improving annotation accuracy and prioritization.
• mWISE is a freely available, high-performing R package for the metabolomics community.