Statistical Techniques in Metabolic Profiling

Metabolic profiling, also known as metabonomics (Nicholson et al., 1999) or metabolomics (Raamsdonk et al., 2001), is a rapidly developing field in biomedical science that combines the application of spectroscopic techniques with multivariate statistical analysis in studies of the molecular composition of biofluids, cells and tissues. The experimental focus is on profiling metabolites the thousands of low molecular weight molecules which are the building blocks of the cell. These compounds interact with macromolecules such as proteins and nucleic acids in the fundamental metabolic processes that keep organisms alive. Thus, the metabolic state of an organism, as delineated by its metabolic profile, can reveal information on the genetic, physiological or functional status the system. In addition, in contrast to DNA or RNA, metabolic profiles also reflect the entirety of internal and external influences on a organism or tissue, including environment, behaviours, disease, drugs and interactions with parasitic or symbiotic organisms (Nicholson et al., 2002; Nicholson and Wilson, 2003). This leads to important applications in the molecular diagnosis and prognosis of disease, drug metabolism and toxicity, functional genomics and the investigation of fundamental biological processes. From the statistical point of view, however, metabolic profiles are complex and information rich, presenting unique challenges to data analysis and modelling. The term metabolome refers to the total metabolite complement of a cell, tissue or organism. Whilst there are currently no methods capable of delivering an exhaustive measurement of the metabolome, there are several technologies that can approach this ideal. A gold standard technique would be one which is a) non-selective, in that it is not biased toward particular chemical classes of compounds, b) quantitative and accurate, thus producing data which can be meaningfully modelled by statistical procedures, c) sensitive to a wide range of metabolite concentrations and d) able to resolve the complex mixture into separate signals from each metabolite. These requirements are satisfied, though not perfectly, by the techniques of nuclear magnetic resonance (NMR) spectroscopy and mass spectrometry (MS).