Computational identification of combinatorial regulation and transcription factor binding sites.

A number of computational methods have been used to unravel the core mechanisms governing the regulation of gene expression, but these techniques examine only portions of the genetic regulatory mechanism. For example, some studies have failed to include the combined action of multiple transcription factors (TFs) or the importance of TF binding constraints (i.e., the binding position and orientation), while others have examined combinations of only 2 or 3 TFs. Thus, we sought to develop a new method for identifying regulatory modules in yeast, using an algorithm that includes all combinations of TFs plus a number of binding constraints when identifying target genes. We successfully developed a computational method for using microarray and TF-DNA interaction data to identify regulatory modules. All possible combinations of yeast TFs and various binding constraints were tested to identify regulatory modules. Within the identified modules, target genes were found to have common binding constraints such as fixed binding regions and orientations for each TF. Moreover, targets showed similar mRNA expression profiles and high functional coherence. Our novel approach, which accounts for both combined actions of TFs and their binding constraints, can be used to identify target genes and reliably predict regulatory modules over a broad range of functional categories. Complete results and additional information are available online at http://bisl. kaist.ac.kr/~dhlee/comModule/index.html.