Figure 6. Diversity of binding properties suggests a layered TF architecture.
A. The TF hierarchy graph. A directed edge goes from a TF X to a TF Y if X binds in at least 30% of the regions bound by Y at the same or earlier time. Edge color is determined by the coverage of X over Y (30–100%; see Table S8 for coverage values). The nodes are color coded according to the percentage of binding sites that were already bound pre-stimulation. For clarity of presentation we employed a pruning strategy (Methods) that removes direct links between nodes at the top of the hierarchy to nodes at the bottom of the hierarchy. Each connected component in the network (rooted at PU.1 or Cebpb) represents a unique combination of TF. The combinations in the trimmed network cover 78% of the TF-region binding data. The number of out-going edges in the non-pruned hierarchy graph (out degree) for nodes at different layers is provided on the left (presented values are the fraction of the maximum out degree). B. Principle Components Analysis was performed with several binding characteristics (Methods). The plot depicts the projections of the TFs and the loading of the different covariates for the first two principle components. C. Model depicting the layered TF network architecture: Pioneer factors initially bind and initiate remodeling of the epigenome, strong binders prime targets for expression and specific TFs control expression of smaller subsets of genes.