Statistical Methods for Handling Unwanted Variation in Metabolomics Data

Abstract

Metabolomics experiments are inevitably subject to a component of unwanted variation, due to factors such as batch effects, long runs of samples, and confounding biological variation. Although the removal of this unwanted variation is a vital step in the analysis of metabolomics data, it is considered a gray area in which there is a recognised need to develop a better understanding of the procedures and statistical methods required to achieve statistically relevant optimal biological outcomes. In this paper, we discuss the causes of unwanted variation in metabolomics experiments, review commonly used metabolomics approaches for handling this unwanted variation, and present a statistical approach for the removal of unwanted variation to obtain normalized metabolomics data. The advantages and performance of the approach relative to several widely-used metabolomics normalization approaches are illustrated through two metabolomics studies, and recommendations are provided for choosing and assessing the most suitable normalization method for a given metabolomics experiment. Software for the approach is made freely available online.

miRNAs Differentially Expressed by Next-Generation Sequencing in Cord Blood Buffy Coat Samples of Boys and Girls

“…We tested quantile normalization in addition to four other common normalization techniques and found that full quantile normalization followed by RUV normalization produced the most optimal results. RUV has been successfully applied in the analysis of large datasets derived from RNA-seq [79] and metabolomic [80] experiments [54]. Our filtering and normalization pipeline resulted in minimization of technical variability, and enabled us to identify biological variability of miRNA expression in boys and girls.…”

Normalization and integration of large-scale metabolomics data using support vector regression

“…The unwanted variations in the measurements of metabolite ion peaks during data acquisition (intra-and inter-batch) are unavoidable and arise from sample handling and preparation, LC column degradation, matrix effects, MS instrument contamination and nonlinear drift over long runs (Leek et al 2010;Burton et al 2008;De Livera et al 2015). Therefore, the development of a normalization method is necessary to remove the unwanted analytical variations occurring in intra-and inter-batch measurements and to integrate multiple batches forming an integral data set for subsequent statistical analysis (De Livera et al 2015;De Livera et al 2012). Effective removal of unwanted analytical variations helps increase the power of statistical analysis so that subtle metabolic changes in epidemiological studies can be detected (Veselkov et al 2011).…”

“…But these scalar correction based methods are not applicable to most metabolomics experiments, as they heavily rely on the selfaveraging property (Sysi-Aho et al 2007). The pros and cons of all normalization methods, including internal standard based normalization, sample-wises scalar normalization, and variance based normalization (De Livera et al 2012;Huber et al 2002), have been comprehensively discussed and compared in several recent articles (van den Berg et al 2006;Kamleh et al 2012;De Livera et al 2015).…”

Metabolite profiles evaluated, according to sex, do not predict resting energy expenditure and lean body mass in healthy non-obese subjects

“…Signal intensity drift, i.e., intra-and interbatch effects occurring during the 4-5 week measurement period were corrected by means of regularly injected quality control (QC) samples [38][39][40]. For the data of the semitargeted GC-MS analysis of sugar species in urine, an automatic method for integration was prepared using the Postrun Analysis feature of GCMSSolution (v 4.1.1.).…”