Statistical spectroscopic methods in drug metabolism and metabolic profiling

Statistical methods for helping characterise and make sense of spectroscopic data are of growing importance in biological research. The quantity of data describing biological objects and systems produced on a daily basis continues to increase, going hand-in-hand with developments in high-throughput techniques, assay multiplexing, minaturisation using nanofluidic devices, and, more commonly, with improvements in instrumental resolution and sensitivity. Trying to organize and interrogate this data to extract the information contained within is therefore a task of increasing complexity, and there is a need to develop methods that efficiently combine data for rapid analysis to help understand and interpret chemical and biochemical processes. Thus, the use of efficient statistical techniques to integrate and directly investigate spectroscopic data has gained prominence in recent years. A number of statistical techniques have been developed from the basis of methods for generalized two-dimensional correlation spectroscopy, summarised in (1). In metabonomics, the most commonly used have been statistical total correlation spectroscopy (STOCSY), statistical heterospectroscopy (SHY), and related techniques for the interpretation of spectroscopic data, which has been demonstrated previously (2-11), and can be applied with relative ease to datasets containing a sufficient number of high quality spectra. As described in Cloarec et al. (2005) (2), central to the STOCSY approach is the calculation of the correlation matrix C that can be obtained from n set(s) of spectra with v spectral variables: