An emergent self-organizing map based analysis pipeline for comparative metabolome studies.

Area of Science:

• Metabolomics
• Bioinformatics
• Computational Biology

Background:

• High-throughput techniques generate vast amounts of complex metabolomic data.
• Efficient computational methods are crucial for interpreting this data.
• Existing methods may lack the ability to fully unravel complex biological system structures.

Purpose of the Study:

• To present a novel computational pipeline for analyzing high-throughput metabolomic data.
• To develop a method for identifying metabolic biomarkers and visualizing pathway differences.
• To validate the pipeline using time-resolved metabolomic datasets.

Main Methods:

• A data analysis pipeline combining Principal Component Analysis (PCA), Emergent Self-Organizing Maps (ESOM), and Hierarchical Cluster Analysis (HCA).
• Utilizing KEGG metabolic pathway maps for automated visualization of metabolomic differences.
• Implementation in user-friendly Java software named eSOMet.

Main Results:

• The pipeline successfully unraveled the structure of complex metabolomic datasets, including outlier detection.
• Automated mapping and visualization using KEGG pathways highlighted metabolic differences across conditions.
• Identified typical metabolic biomarkers for different carbon sources and growth phases in Corynebacterium glutamicum.

Conclusions:

• The presented computational pipeline is effective for analyzing high-throughput metabolomic data.
• The eSOMet software provides a user-friendly tool for metabolomic data clustering and biomarker discovery.
• This approach facilitates the identification of condition-specific metabolic biomarkers and pathway alterations.