Supervised multivariate statistical analyses of NMR spectroscopic data sets are often required to identify metabolic differences between sample classes, and the use of orthogonal filters has proven to be highly efficient even when dealing with weak perturbations. In this note, we associate orthogonal filters to the recently reported recoupled-statistical total correlation spectroscopy (RSTOCSY). An initial supervised deflation of the spectral matrix is applied to remove all information orthogonal to the effect of interest and is followed by an RSTOCSY analysis to extract a list of pairs of metabolites that experience correlated perturbations. This list can then be used to find possibilities for the perturbed metabolic network. This supervised RSTOCSY approach, dubbed OR-STOCSY, yields metabolites related to perturbations of biological interest, even if they make a minor contribution to the global variance of a complex data set compared to other (possibly confounding) effects under study. The method is demonstrated with the application to genetic phenotypes in Caenorhabditis elegans.