SRV: an open-source toolbox to accelerate the recovery of metabolic biomarkers and correlations from metabolic phenotyping datasets

摘要

Motivation: Supervised multivariate statistical analyses are often required to analyze the high-density spectral information in metabolic datasets acquired from complex mixtures in metabolic phenotyping studies. Here we present an implementation of the SRV—Statistical Recoupling of Variables—algorithm as an open-source Matlab and GNU Octave toolbox. SRV allows the identification of similarity between consecutive variables resulting from the high-resolution bucketing. Similar variables are gatheredto restore the spectral dependency within the datasets and identify metabolic NMR signals. The correlation and significance of these new NMR variables for a given effect under study can then be measured and represented on a loading plot to allow a visualand efficient identification of candidate biomarkers. Further on, correlations between these candidate biomarkers can be visualized on a two-dimensional pseudospectrum, representing a correlation map, helping to understand the modifications of the underlying metabolic network. Availability: SRV toolbox is encoded in MATLAB R2008A (Mathworks, Natick, MA) and in GNU Octave. It is available free of charge at http:// www.prabi.fr/redmine/projects/srv/repository with a tutorial.