Abstract
Deciphering the modularity of transcriptional networks is a principle approach to understand this complex biological system. We purpose a multi-step clustering scheme to extract sets of genes regulated by the same set of transcription factors. These sets of genes are defined as transcriptional modules. In our approach, we first obtain significant evidences of co-expression from multiple microarray time profiles, and then extract sets of genes which forms dense sub-graph on the co-expression networks by a novel method called Clique-based Dense Sub-graph Finding Algorithm. We apply this scheme to artificial data and real microarray datasets of Saccharomyces cerevisiae. The resulted modules can potently imply the topological structure of transcriptional networks, and also have significant annotated function, component, or process. In the promoter sequence analysis, we can also find significant binding motifs, and have agreement on function between resulted modules and known motifs.