Extraction, interpretation and validation of information for comparing samples in metabolic LC/MS data sets

Area of Science:

• Analytical Chemistry
• Metabolomics
• Systems Biology

Background:

• Liquid chromatography-mass spectrometry (LC/MS) is a sensitive technique for metabolic profiling.
• Generating robust and interpretable multivariate models requires specific data criteria for accurate biological interpretation.
• Existing methods may not fully address the challenges of large-scale LC/MS data analysis for comparative studies.

Purpose of the Study:

• To develop and validate a generalizable data handling, analysis, and evaluation strategy for large metabolic LC/MS datasets.
• To create robust, interpretable, and predictive multivariate models for comparing biological samples.
• To identify metabolic differences between distinct populations using LC/MS data.

Main Methods:

• Data preprocessing including automatic peak detection, alignment, and retention time window setting.
• Data compression using alternating regression to retain relevant metabolic variation.
• Multivariate analysis, specifically partial least square discriminant analysis (PLS-DA), for model building and prediction.

Main Results:

• The developed strategy successfully processed and analyzed LC/MS data from urine samples of two distinct populations (Shanxi, China and Honolulu, USA).
• Partial least square discriminant analysis (PLS-DA) yielded a robust, predictive, and transparent model for differentiating metabolic profiles.
• The approach enabled identification of systematic patterns and influential variables, facilitating biological interpretation.

Conclusions:

• The presented strategy offers a generalizable approach for handling and analyzing large-scale metabolic LC/MS data.
• The method facilitates the creation of reliable and interpretable models for comparative metabolomics.
• This strategy is effective for identifying and understanding metabolic variations between different populations or conditions.