Mixture model normalization for non-targeted gas chromatography/mass spectrometry metabolomics data

Area of Science:

• Biochemistry
• Analytical Chemistry
• Computational Biology

Background:

• Metabolomics provides an integrated view of health, reflecting genetic and environmental factors.
• Large-scale studies require batch processing for gas-chromatography/mass spectrometry (GC/MS) assays, introducing technical noise.
• Existing normalization methods struggle with batch effects and truncated low-abundance compounds in metabolomics data.

Purpose of the Study:

• To develop a novel normalization method for GC/MS metabolomics data.
• To address technical noise, batch effects, and data truncation in non-targeted assays.
• To improve the interpretability and biological relevance of metabolomics data.

Main Methods:

• Proposed mixture model normalization (mixnorm) to handle truncated data.
• Utilized quality control samples to estimate per-metabolite batch and run-order effects.
• Compared mixnorm performance against existing normalization approaches.

Main Results:

• Mixnorm demonstrated superior performance across multiple metrics compared to other methods.
• Achieved improved correlation between non-targeted and targeted metabolomics measurements.
• Showcased enhanced performance when metabolite detectability varied by batch.

Conclusions:

• Mixnorm is uniquely suited for normalizing non-targeted GC/MS metabolomics data, especially in large studies.
• The method explicitly accommodates batch effects, run order, and varying detection thresholds.
• Accurate normalization is critical for reliable conclusions from large-scale GC/MS metabolomics studies.