Analysis and Interpretation of Metabolite Associations Using Correlations

In metabolomic studies, metabolite concentrations and their associative relationships are usually quantified and interpreted in term of correlation indexes like the Pearson’s and Spearman’s correlation coefficients. The aim of such analysis is usually to obtain information about the underlying metabolic network and its characteristics. However, there is often no direct relationship between the patterns of metabolite associations and the structure of the metabolic networks regulating metabolite concentrations. The first part of the chapter reviews the basic biochemical mechanisms from which the patterns of correlations observed in metabolomic data arise with a focus on their interpretation. The second part reviews the most common measures of association (Pearson’s, Spearman’s, and Kendall’s correlation coefficient and mutual information) and the statistical assumption underlying their applicability. Robust methods are briefly discussed, and a component model for the comprehensive analysis of correlation matrices is also presented.