Comparison of standardization approaches applied to metabolomics data

Abstract

Some factors such as unwanted variations might affect the identification of biomarkers in metabolomics and proteomics analysis, which needs preprocessing including normalization (also named as standardization) by the standardization approach prior to marker selection. Many standardization approaches were applied to analysis of the metabolomics, and even proteomics data. But there are rarely comprehensive comparison of the standardization performance based on the sample size and various methods. The current study performed an overall comparison aiming at these methods based on a metabolomics dataset. As a result, 15 standardization approaches were classified into four groups according to the standardization performances of different sample sizes. The Log Transformation and the VSN method were regarded as the Superior performance methods, but the Contrast method was performed consistently worst in all datasets of various sample size. This study could provide a useful guidance for the choice of befitting standardization approaches when carrying out the metabolomics and proteomics analysis based on LC/MS.