ASCA: The Implementation of Design of Experiments Into Multivariate Modelling in Chemometrics

Abstract The untargeted metabolomics paradigm may be very helpful to reveal biochemical patterns in the multifactorial experiments in environmental analysis. Specific combinations between metabolites may be highly specific biomarkers for environmental and/or ecological change. However, such patterns need to be recovered from a background of many unrelated metabolites within a wealth of cooccurring environmental processes. The range of methods we present here, based around Analysis-of-Variance-Simultaneous Component Analysis (ASCA), have been specifically developed retrieve such patterns. They combine the merits of quantitatively describing the Design-of-Experiments that underlies an environmental study with the multivariate nature of metabolomics data. The ASCA toolbox has by now extended into a comprehensive and generic approach that allows analysis of, e.g., unbalanced data, quantitative significance of effects and of relevant biomarkers. The same approach can also be taken to analyse specific effects with respect to positive and negative controls, to reveal specific experimentally relevant deviations in metabolism. We show the ASCA results of a specific plant chemical ecology dataset, in which all glucosinolates within a wild cabbage were profiled upon induction of an ecological defence response. ASCA provides direct insight in the variability associated with different aspects of this response, and relatively recent extensions in data preprocessing reveal very clearly the metabolites that are most relevant to the response.