Equating, or Correction for Between-Block Effects with Application to Body Fluid LC−MS and NMR Metabolomics Data Sets

摘要

Combination of data sets from different objects (fornexample, from two groups of healthy volunteers from thensame population) that were measured on a common setnof variables (for example, metabolites or peptides) isndesirable for statistical analysis in “omics” studies becausenit increases power. However, this type of combinationnis not directly possible if nonbiological systematicndifferences exist among the individual data sets, orn“blocks”. Such differences can, for example, be due tonsmall analytical changes that are likely to accumulate overnlarge time intervals between blocks of measurements. Innthis article we present a data transformation method, thatnwe will refer to as “quantile equating”, which per variablencorrects for linear and nonlinear differences in distributionnamong blocks of semiquantitative data obtained withnthe same analytical method. We demonstrate the successfulnapplication of the quantile equating method to datanobtained on two typical metabolomics platforms, i.e.,nliquid chromatography-mass spectrometry and nuclearnmagnetic resonance spectroscopy. We suggest uni- andnmultivariate methods to evaluate similarities and differencesnamong data blocks before and after quantile equating.nIn conclusion, we have developed a method to correctnfor nonbiological systematic differences among semiquantitativendata blocks and have demonstrated its successfulnapplication to metabolomics data sets.