Assessing normalization methods in mass spectrometry-based proteome profiling of clinical samples.

Area of Science:

• Proteomics
• Bioinformatics
• Clinical Research

Background:

• Large-scale proteomic studies face significant variability from sample collection to data analysis.
• Normalization is essential to mitigate this variability in multi-center, multi-batch clinical research.
• Few reviews focus on normalization for mass spectrometry (MS)-based proteomics, especially for large datasets.

Purpose of the Study:

• To evaluate various normalization methods for large-scale MS-based proteomic data.
• To assess the impact of normalization on the relationships between proteins and clinical variables.
• To identify optimal normalization strategies for human plasma proteomic datasets.

Main Methods:

• Applied multiple normalization techniques including quantile sample, RUV, mean/median centering, and ComBat.
• Utilized a large proteomic dataset from an overweight and obese pan-European cohort.
• Analyzed improvements in protein-clinical variable associations post-normalization.

Main Results:

• Quantile sample normalization, RUV, mean centering, and median centering demonstrated strong performance.
• Quantile protein normalization yielded poorer results compared to unnormalized data.
• Normalization improved the detection of relationships between proteins and clinical factors like gender and lipid levels.

Conclusions:

• Specific normalization methods are better suited for large-scale shotgun proteomic data from human plasma.
• Quantile sample normalization, RUV, mean, and median centering are recommended for such datasets.
• Quantile protein normalization is not advised for this type of proteomic data.