By the company they keep: interaction networks define the binding ability of transcription factors

Access to genome-wide data provides the opportu-nity to address questions concerning the ability of transcription factors (TFs) to assemble in distinct macromolecular complexes. Here, we introduce the PAnDA (Protein And DNA Associations) approach to characterize DNA associations with human TFs using expression profiles, protein–protein interac-tions and recognition motifs. Our method predicts TF binding events with>0.80 accuracy revealing cell-specific regulatory patterns that can be exploited for future investigations. Even when the precise DNA-binding motifs of a specific TF are not available, the information derived from protein-protein net-works is sufficient to perform high-confidence pre-dictions (area under the ROC curve of 0.89). PAnDA is freely available a