Schematic of Overall Data Analysis Approach. Gene chip data were analyzed using the RMA probe set algorithm. We then used the Statistical Analysis of Microarray (SAM) approach with control of false discovery rate to identify 1484 differentially expressed genes across our experimental groups. Clustering analysis was then performed; promoters of genes from clusters of interest were then identified and TFBS enrichment analysis was performed. This was accomplished with PWMs from Transfac and the CLOVER algorithm. For estimation of the background frequency of TFBSs, we compared our experimental data to thousands of random promoter sequences. We also performed enrichment analysis at the gene level, which told us if certain TFBSs were present in more of our experimental promoters than in random samplings of background promoters. These analyses led to the identification of enriched TFBSs that were conserved in rat, mouse and human. The final analysis allowed us to predict synergistic interactions between TFs by estimating the frequency of the presence of TFBS combinations in our experimental promoters as compared to random, background promoter sequences.
Collins and Hu BMC Genomics 2007 8:420 doi:10.1186/1471-2164-8-420