mWISE

mWISE performs context-based annotation of LC-MS metabolomics data to prioritize and assign KEGG metabolite candidates for untargeted metabolomics studies.

Key Features:

• Context-Based Annotation: Leverages contextual information to increase confidence of metabolite annotations in untargeted LC-MS metabolomics.
• Mass-to-Charge Ratio Matching: Matches mass-to-charge ratio (m/z) values from LC-MS data to entries in the Kyoto Encyclopedia of Genes and Genomes (KEGG).
• Clustering and Filtering: Clusters and filters potential KEGG candidates to refine and reduce the list of possible annotations.
• Diffusion-Based Prioritization: Employs a diffusion process on a biological network derived from the FELLA R package combined with raw scores to build a prioritized list of candidate metabolites.
• Evaluation on Public Datasets: Evaluated using three publicly available studies that include both positive and negative ionization modes.
• Benchmarking against xMSannotator: Benchmarked across four configurations versus xMSannotator, reporting superior sensitivity and computation time.
• Sensitivity and Structure Accuracy: Reported average sensitivity 0.63 (SD 0.07) compared to xMSannotator 0.55 (SD 0.19), and proposed chemical structures were closer to the original compounds.
• Diffusion Prioritization Role: Demonstrated that diffusion prioritization substantially contributes to improved annotation accuracy.

Scientific Applications:

• Untargeted metabolomics annotation: Improves identification and confidence of metabolite assignments from LC-MS experiments.
• Metabolic pathway analysis: Facilitates mapping of annotated metabolites to biological pathways using KEGG-derived candidates.
• Systems biology: Supports generation of network-contextualized metabolite hypotheses for integrative analyses.
• Pharmacology and biomarker discovery: Aids in prioritizing candidate metabolites relevant to drug response and biomarker identification.
• Personalized medicine: Enables more confident metabolite identification that can inform individualized metabolic profiling.

Methodology:

Computational steps explicitly include m/z matching of LC-MS features to KEGG entries, clustering and filtering of KEGG candidates, and a diffusion process on a biological network derived from the FELLA R package combined with raw scores to produce a prioritized list of candidate metabolites; evaluation used three public studies with positive and negative ionization modes and benchmarking against xMSannotator across four configurations.