A data preprocessing strategy for metabolomics to reduce the mask effect in data analysis.

Area of Science:

• Metabolomics
• Data Analysis
• Biochemistry

Background:

• Metabolomics research relies on identifying differential metabolites between sample groups.
• High variation in abundant metabolites can mask subtle differences in low abundant ones.
• Existing methods struggle with missing values and measurement deviations in metabolomics data.

Purpose of the Study:

• To develop a robust data preprocessing strategy for metabolomics.
• To address challenges posed by missing values and mask effects.
• To enhance the discovery of low abundant differential metabolites.

Main Methods:

• A three-step preprocessing strategy was developed.
• Step 1: Modified 80% rule for missing value imputation.
• Steps 2 & 3: Unit-variance and Pareto scaling, followed by stability-weighted adjustments to mitigate mask effects and scaling deviations.
• A new method, 'x-VAST', was introduced to correct measurement deviation enlargement.

Main Results:

• The proposed strategy effectively reduced mask effects in metabolomics data.
• Several low abundant differential metabolites, previously masked, were successfully identified.
• The method demonstrated efficacy on a chronic hepatitis B patient dataset and simulated datasets.

Conclusions:

• The novel preprocessing strategy significantly improves the detection of differential metabolites.
• This approach is crucial for advancing metabolomics research, particularly in disease biomarker discovery.
• The 'x-VAST' method and preprocessing pipeline offer a valuable tool for analyzing complex biological data.