Abstract
To understand the transcriptional regulatory network in eukaryotic cells, it is essential to identify functional cis-regulatory sequences that interact with trans-acting factors. A number of algorithms have been developed to predict common cis-regulatory elements for co-regulated genes with similar expression patterns. However, previous methods usually deal with disjoint gene groups partitioned or clustered by arbitrary cutoffs, which might cause information losses. To preclude the defining step of gene set, we adopted enrichment analysis and termed the method binding site enrichment analysis (BSEA). BSEA was first applied for publicly available ChIP-on-chip data of c-MYC, MAX and E2F transcription factors, identifying significant enrichment for signatures of corresponding factors and potential co-activators. Using time-scaled expression profiling of 3T3-L1 adipogenesis, we observed enrichment for signatures of known adipogenic factors such as C/EBPα, C/EBPβ and PPARγ, temporally coincident with previous reports. BSEA was also applied to tissue-specific expression profiles of human and mouse, identifying well-known tissue-specific transcription factors such as HNF-4 in liver and MEF-2 in heart along with other putative tissue-specific regulators. With extended versatility coping with various kinds of microarray dataset, BSEA can identify key regulators for global microarray data in which transcriptional regulation plays a major role. As a generalized method, BSEA would help to elucidate the transcriptional regulatory networks, the primary challenges in functional genomics.