Computational identification of transcription factors involved in early cellular response to a stimulus

The response of genes to cell stimuli is often measured by microarrays. However, studying the function of these genes rarely elucidate as to how the stimuli activate or suppress these genes. To understand the mechanisms of cell stimulation, we describe a computational method for analyzing mammalian promoters of early response genes to detect the transcription factors activated by cell stimulation. We first analyzed promoters of the response genes, for transcription factor binding sites conserved between human and mouse. We then applied hypergeometric statistics in conjunction with Bonferroni correction to identify the top transcription factors whose binding sites were significantly over-represented among these promoters. In five data sets with early response genes, a significantly larger than expected number of genes had binding sites in their promoters for transcription factors previously known to be involved in response to the stimulus, while data sets with measurements at longer time points (24 hours) failed to show such over-representation. Because the end points of signal transduction pathways are transcription factors, this methodology is useful for exploring signaling pathways activated by various stimuli through microarray studies.