Extended Data Fig. 2: Integrative modeling to quantify TF binding cooperativity.

(a) Schematic table describing the combinations of TFs assayed in five experiments (top) that were jointly analyzed to produce binding models of the different monomers and their complexes (bottom) by explicitly defining which models can form in each experiment (+ sign). (b) Distribution of probes (top) and the predicted relative contribution of every recognition mode (bottom) as a function of predicted binding selection strength (x-axis) in the first round of selection from SELEX-seq data assaying Hth, Exd, and UbxIV. (c) Integrative modeling of HT-SELEX and CAP-SELEX data for MEIS1 and DLX3 (schematic table) yields binding models for the monomers (energy logos) and configuration-dependent binding cooperativity for the MEIS1:DLX3 complex (same circle plot representation as in Fig. 3b). The bottom right logo shows the specificity of MEIS1:DLX3 for the most stable configuration (connecting arrow), aligned to a sequence previously crystallized with MEIS1:DLX31. (d) Table showing the availability of CAP-SELEX data for different TF-TF combinations. The 10 TFs with the most identified co-factors are included, and numbers indicate replicate count. (e) Distribution plot comparing the binding cooperativity inferred by ProBound at the configurations that were identified as cooperative in the original CAP-SELEX study (red line) and at all other configurations (gray line). The models were trained on the CAP-SELEX data tabulated in (d) and are shown in Extended Data Figure 3.