Multiway chemometric analysis of the metabolic response to toxins monitored by NMR

For the first time, the multiway chemometric model Tucker3 was applied to nuclear magnetic resonance (NMR) time series data from studies of the metabolic response to toxins. Three data sets were analysed, each consisting of ID high-field H-1 NMR spectra of rat urine samples collected over 7 days after administration of a single dose of a model toxin in different doses. The analysis using the multiway Tucker3 model resulted in models with two to three components in each mode. The observed perturbations following administration of the toxins corresponded well to those reported in other studies using two-way chemometric methods (PCA and batch processing analysis). The advantage of the Tucker3 analysis is that the three-way structure of the data (rat, time, reduced NMR spectra) is retained and used throughout the analysis. Only the metabolic perturbations that have common time profiles are modelled, whereas random or less general changes in metabolite levels, which do not have common time profiles, are considered as noise. In the Tucker3 model, no constraints are imposed on time profiles, and thus, the models can handle effects with complex time evolution behavior. (c) 2004 Elsevier B.V. All rights reserved.