Unbiased and biased chemometric analysis of LC‐MS data from human urine following coffee intake

Abstract

AbstractWe carried out a human volunteer study with 14 participants, eight of whom were asked to consume one cup of coffee at four different time points. Urine samples were collected at eight time points and analyzed by HPLC‐MS analysis. The LC‐MS data were subjected to unsupervised multivariate statistical analysis (principal component analysis) followed by supervised multivariate analysis (linear discriminant analysis). In an unbiased approach, in the absence of data preselection and filtering, the most important features explaining differences between coffee consumers and the control group observed showed variations in endogenous human hormonal steroid metabolites as well as xanthine derivatives. Only after a biased data treatment data revealed differences between the sample groups based on literature reported chlorogenic acid metabolites resulting directly from coffee intake. Such analysis could confirm the presence of 21 previously reported chlorogenic acid plasma metabolites as urinary metabolites. The application of tandem MS molecular networking revealed the presence of five bioavailable chlorogenic acid derivatives in urine previously not reported, including both quinic acid lactone and dimethoxy caffeoyl esters. Selected cinnamic acids were quantified in urine.