In the post-genomic era, public databases like GEO and NASCArrays provide a large volume of array data. These array data of various experimental conditions give scientists a base for computing co-expression gene modules. In order to study iron(Fe)-responsive and phosphate(P)-responsive genes in Arabidopsis roots, we developed the MACCU toolbox (http://maccu.sourceforge.net/) for co-expression computation. In addition to computing co-expression networks based on publicly available array data, we used MACCU to compare co-expression networks made based on different sets of array data. Accordingly, we deduced root-specific co-expression networks with reduced biological complexity, which help biologists narrow down interesting responsive genes.