Fig. 5. ProBound learns quantitative binding models and sample-specific TF activities using peak-free ChIP-seq analysis.

a, Binding models for GR and three co-factors (left) learned from GR ChIP-seq data from the IMR90 cell line⁴⁷ and for GR from a SELEX dataset (center). The scatterplot compares the energy coefficients learned from ChIP-seq (y axis) and SELEX (x axis) data⁷. b, Combined specificity (top) and sample-specific TF binding activity (bottom) model learned by jointly analyzing three GR ChIP-seq datasets after treatment with 30 μg kg⁻¹, 300 μg kg⁻¹ or 3,000 μg kg⁻¹ of CORT⁵¹. The scatterplot (left) compares the energy coefficients as in a.