Microarray data produces expression pattern of thousands of genes at once. Grouping these gene expression patterns to have each group convey some biologically meaningful sight entails use of a clustering method. Two problems exist when attempting to use conventional clustering methods for the microarray data analysis. Presence of outliers skews the mean value computation which, in turn influences placement of inconsistent gene expression patterns into one group. The clustering algorithms themselves generally cannot determine the right size of the clusters. We present a new method which approaches to the clustering problem from a different angle. That is, the clustering of gene expression patterns is better dealt with within a software framework that is conducive to helping biologists derive the right size of clusters utilizing their understanding of the experimental context once the baseline clusters are computed using the fold changes of gene expression levels. We discuss our experiences of using the framework in analyzing numerous microarray data experiments.