Evaluation of normalization methods to pave the way towards large-scale LC-MS-based metabolomics profiling...

Area of Science:

• Metabolomics
• Analytical Chemistry
• Bioinformatics

Background:

• Combining liquid chromatography-mass spectrometry (LC-MS) metabolomics data from long-term experiments is challenging due to systematic variability.
• Existing model-driven normalization methods often rely on internal standards or external models, limiting their applicability.

Purpose of the Study:

• To evaluate data-driven normalization approaches for LC-MS metabolomics experiments.
• To assess the effectiveness of cyclic-Loess normalization for removing systematic variability over time.
• To recommend optimal normalization strategies for untargeted LC-MS metabolomics.

Main Methods:

• Evaluation of existing data-driven normalization methods on LC-MS metabolomics datasets.
• Application of cyclic-Loess normalization to a Leishmania sample dataset.
• Analysis of variability measures and identification of differential metabolites.

Main Results:

• Data-driven normalization methods significantly improve data analysis from multiple experimental runs.
• Cyclic-Loess normalization successfully removes systematic variability between measurement blocks while preserving differential metabolites.
• All evaluated normalization methods demonstrated improved data analysis.

Conclusions:

• Normalization is crucial for pooling LC-MS metabolomics datasets, enhancing statistical power and enabling larger-scale experiments.
• Data-driven normalization methods are recommended over model-driven methods, especially for untargeted LC-MS experiments.
• Normalization strategies are essential for robust and scalable metabolomics research.