Predicting transcription factor synergism

Transcriptional regulation is mediated by a battery of transcription factor (TF) proteins, that form com-plexes involving protein±protein and protein±DNA interactions. Individual TFs bind to their cognate cis-elements or transcription factor-binding sites (TFBS). TFBS are organized on the DNA proximal to the gene in groups con®ned to a few hundred base pair regions. These groups are referred to as modules. Various modules work together to provide the combinatorial regulation of gene transcription in response to various developmental and environ-mental conditions. The sets of modules constitute a promoter model. Determining the TFs that preferen-tially work in concert as part of a module is an essential component of understanding transcrip-tional regulation. The TFs that act synergistically in such a fashion are likely to have their cis-elements co-localized on the genome at speci®c distances apart. We exploit this notion to predict TF pairs that are likely to be part of a transcriptional module on the human genome sequence. The computational method is validated statistically, using known inter-acting pairs extracted from the literature. There are 251 TFBS pairs up to 50 bp apart and 70 TFBS pairs up to 200 bp apart that score higher than any of the known synergistic pairs. Further investigation of 50 pairs randomly selected from each of these two sets using PubMed queries provided additional support-ing evidence from the existing biological literature suggesting TF synergism for these novel pairs