In modern biology, we had an explosion of genomic data from multiple sources, like measurements of RNA levels, gene sequences, annotations or interaction data. These heterogeneous data provide important information that should be integrated through suitable learning methods aimed at elucidating regulatory networks. We propose an iterative relational clustering procedure for finding modules of co-regulated genes. This approach integrates information concerning known Transcription Factors (TFs)--gene interactions with gene expression data to find clusters of genes that share a common regulatory program. The results obtained on two well-known gene expression data sets from Saccharomyces cerevisiae are shown.