An efficient algorithm to identify coordinately activated transcription factors.

Identification of transcription factor (TF) activities associated with a certain physiological/experimental condition is one of the preliminary steps to reconstruct transcriptional regulatory networks and to identify signal transduction pathways. TF activities are often indicated by the activities of its target genes. Existing studies on identifying TF activities through target genes usually assume the equivalence between co-regulation and co-expression. However, genes with correlated expression profiles may not be co-regulated. In the mean time, although multiple TFs can be activated coordinately, there is a lack of efficient methods to identify coordinately activated TFs. In this paper, we propose an efficient algorithm embedding a dynamic programming procedure to identify a subset of TFs that are potentially coordinately activated under a given condition by utilizing ranked lists of differentially expressed target genes. Applying our algorithm to microarray expression data sets for a number of diseases, our approach found subsets of TFs that are highly likely associated with the given disease processes.