The effects of sample preparation and chromatographic method differences on the classification and recovery of metabolic biomarkers from UPLC-MS measurements on urine samples of humans exposed to different dietary interventions have been investigated. Eight volunteers consumed three high-fat meals (rich in saturated, monounsaturated and polyunsaturated fatty acids, respectively) in randomized order with a washout period in between. For each participant, urine samples were obtained prior to and at three timed intervals after each meal. Samples were processed either by dilution (1 : 4) or by liquid-liquid extraction and then run under two different gradient conditions. For each analysis method, a total of 96 observations (eight participants, four time points, three diets) were measured. The total ion count chromatograms were analyzed using partial-least-squares discriminant analysis. All three dietary classes could be discriminated irrespective of sample preparation and chromatographic method. However, the main discriminating metabolites varied according to sample preparation, indicating that sample treatment and chromatographic conditions influence the ability to extract biomolecular information. Diluted samples showed higher m/z compounds (ca 400 u) while liquid-liquid extraction samples showed low m/z at the same retention time span. Optimized methods for metabolite identification (e.g. organic acids) were statistically inferior to global screening for mixed compound identification, confirming that multiple compound class-based metabolic profiles are likely to give superior metabonomic (diagnostic) classification, although great care has to be taken in the interpretation in relation to matrix effects.
Copyright (c) 2009 John Wiley & Sons, Ltd.