Network Marker Selection for Untargeted LC-MS Metabolomics Data.

Area of Science:

• High-throughput biology
• Metabolomics
• Systems biology

Background:

• Untargeted metabolomics via high-resolution liquid chromatography-mass spectrometry (LC-MS) is crucial for biological research.
• Functional analysis of metabolomics data is vital but challenged by feature identification uncertainty and missing measurements.
• Existing methods often overlook identification ambiguities and missing data, potentially leading to inaccurate pathway analysis.

Purpose of the Study:

• To develop a flexible framework for network feature selection integrating metabolomics data with genome-scale metabolic networks.
• To simultaneously address metabolite identification uncertainty and missing observations in functional analyses.
• To improve the accuracy and sensitivity of selecting significant biological subnetworks.

Main Methods:

• A novel framework combining metabolomics data with genome-scale metabolic networks.
• Sequential feature screening procedure coupled with machine learning criteria for optimal feature matching.
• Simultaneous selection of important subnetworks and identification of the true metabolite match.

Main Results:

• The proposed method demonstrates significantly higher sensitivity compared to the maximal matching approach in simulation studies.
• Application to a healthy cohort identified subnetworks associated with body mass index (BMI).
• Identified subnetworks are supported by existing literature and suggest novel functional implications.

Conclusions:

• The developed framework offers a robust approach to handle uncertainties in metabolomics data analysis.
• It enhances the reliability of functional analysis and subnetwork identification in high-throughput biology.
• The method provides valuable insights into biological systems, with potential for new discoveries.