Fig. 1: Large-scale mapping of E. coli TF binding regions.

A Summary of novel regions (green) and known regions (found:blue, not found:red) for 139 mapped TFs, ordered by total number of binding regions (TFs in blue have BoltzNet models). B Mode of TF expression used to map TFs. Both experimental approaches required to map all TFs. Recovery of known sites by different expression modes for TFs mapped by (C) either or both modes of expression, or (D) both modes only. No single approach recovers all known sites, although TF induction performs better. E Example for PdhR of comparisons of ChIP-Seq experiments within and between expression modes. Axes plot log₁₀ peak enrichment with green circles called in both replicates; orange called only in the Y-axis experiment; and blue called only in the X-axis experiment. Points with a black outline have been previously reported. F Correlation of enrichment across all replicates for all TFs. G Location of binding regions relative to start position of nearest gene (G – genic, IG – intergenic). Binding regions between 150 bp upstream and 50 bp downstream are overrepressented relative to random expectation (grey circles are means with black error bars displaying standard deviation of 100 random samples), regions > 1.5 kb up or down-stream of the nearest gene are grouped into a single bar on either edge. Data for regions with known sites shown in inset. H Both novel and known regions are observed over 3 logs of enrichment. Each row is a TF and each dot is a called region (green new region, blue known region).