Fig. 1

A predictive model of active enhancers. a Feature selection using H3K27ac (minus the Input DNA) as a correlation class. The bars represent the average importance score per feature. Red labels and bars indicate the minimum (randMin), mean (randMean) and maximum (randMax) of the simulated replicates, as well as the ChIP-Seq with a non-specific antibody (Control). The red dashed line separates the relevant features (in blue) from the non-relevant features (in grey). b Scatter plot of the intergenic windows according to relative enrichment z-scores for every pair of selected features (x and y axis). Each dot represents a window and windows are separated according to the three classes: active enhancers (green), no-change regions (blue) silent enhancers (red). The black centroids show the centers and standard deviations of the correlations between different features