The canonical correlation analysis (CCA) is commonly used to analyze data sets with paired data, e.g. measurements of gene expression and metabolomic intensities of the same experiments. This allows to find interesting relationships between the data sets, e.g. they can be assigned to biological processes. However, it can be difficult to interpret the processes and often the relationships observed are not related to the experimental design but to some unknown parameters. Here we present an extension of the penalized CCA, the supervised penalized approach (spCCA), where the experimental design is used as a third data set and the correlation of the biological data sets with the design data set is maximized to find interpretable and meaningful canonical variables. The spCCA was successfully tested on a data set of Arabidopsis thaliana with gene expression and metabolite intensity measurements and resulted in eight significant canonical variables and their interpretation. We provide an R-package under the GPL license.