W4M00001_Sacurine-statistics

Study Characterization of the physiological variations of the metabolome in biofluids is critical to understand human physiology and to avoid confounding effects in cohort studies aiming at biomarker discovery. Dataset In this study conducted by the MetaboHUB French Infrastructure for Metabolomics, urine samples from 184 volunteers were analyzed by reversed-phase (C18) ultrahigh performance liquid chromatography (UPLC) coupled to high-resolution mass spectrometry (LTQ-Orbitrap). A total of 258 metabolites were identified at confidence levels provided by the metabolomics standards initiative (MSI) levels 1 or 2. Workflow This history describes the statistical analysis of the data set from the negative ionization mode (113 identified metabolites at MSI levels 1 or 2): correction of signal drift (loess model built on QC pools) and batch effects (two batches), variable filtering (QC coefficent of variation < 30%), normalization by the sample osmolality, log10 transformation, sample filtering (Hotelling, decile and missing pvalues > 0.001) resulting in the HU_096 sample being discarded, univariate hypothesis testing of significant variations with age, BMI, or between genders (FDR < 0.05), and OPLS(-DA) modeling of age, BMI and gender. Comments The ‘sacurine’ data set (after normalization and filtering) is also available in the ropls R package from the Bioconductor repository. For a comprehensive analysis of the dataset (starting from the preprocessing of the raw files and including all detected features in the subsequent steps), please see the companion ‘W4M00002_Sacurine-comprehensive’ reference history.