CHAPTER 2. Experimental Design: Sample Collection, Sample Size, Power Calculations, Essential Assumptions and Univariate Approaches to Metabolomics Analysis

In this chapter, we discuss and assess essential criteria and investigation-specific, essential requirements for biofluid and/or tissue biopsy sample collection, raw data preprocessing stages (1H NMR, LC-MS or otherwise), and dataset normalisation, scaling and dimensionality reduction processes. Moreover, the critical assumptions required for both univariate and consequently also MV statistical evaluations of such datasets are also discussed, as are those for homoscedasticity and, where appropriate, additivity. In addition, we also outline the importance of the implementation of experimental design models to such investigations, and also the application of increasingly complex ANOVA systems to the analysis of metabolomics or genomic datasets in a multivariate context, in the form of ASCA models, which can also incorporate components of variance arising from interactions between two or more factors, i.e. non-additive responses. Further attention is given to the prior applications of univariate model systems to the analysis of MV metabolomics datasets, together with the critical considerations and constraints which must be applied to such systems. Finally, we also discuss critical sample size requirements for both the univariate and MV analyses of experimental datasets (the latter of which is a newly developing research area), together with essential error analysis, and also the concerns associated with such explorations.