How to do quantile normalization correctly for gene expression data analyses.

Area of Science:

• Bioinformatics
• Computational Biology
• Statistical Genetics

Background:

• Quantile normalization is a standard technique for high-dimensional data.
• Blind application is susceptible to class-effect proportion and batch effects, increasing errors.

Purpose of the Study:

• Evaluate strategies for quantile normalization.
• Improve performance by mitigating batch effects and class-effect proportion issues.

Main Methods:

• Compared five quantile normalization strategies.
• Implemented a "Class-specific" approach, normalizing data splits independently.
• Utilized simulations with real and simulated batch effects.

Main Results:

• The "Class-specific" strategy significantly improved batch-effect correction and feature selection.
• This method outperformed whole-data quantile normalization.
• Preserved useful biological signals even when combining separately normalized datasets.

Conclusions:

• "Class-specific" quantile normalization is a robust strategy for high-dimensional data analysis.
• Recommends this approach over whole-data normalization to enhance accuracy and reliability.