Fig. 1: In vivo mapping of E. coli RNA structure ensembles.
a, Schematic representation of the E. coli genome. Regions populating two or more conformations in TOP10 (dark green), DH5α (light green) or both (gray) are indicated. Examples of riboswitches and RNA thermometers known to populate two alternative conformations are shown. For the mgtL Mg2+ sensor and the hisL histidine leader, reconstructed reactivities for the two conformations were averaged across DH5α and TOP10 cells and overlaid on the known structures. Bases falling outside of the region deconvolved by DRACO are marked in pink. Inset, pie chart depicting the percentages of bases in the E. coli transcriptome populating one, two or three or more conformations. Only bases populating the same number of conformations in DH5α and TOP10 cells were considered. b, Distribution of median Shannon entropies (from unconstrained predictions) in 1 versus 2+ regions. c, Distribution of median unpaired probabilities (from unconstrained predictions) in 1 versus 2+ regions. d, Distribution of median bulk (ensemble average) in vivo DMS reactivities in 1 versus 2+ regions. e, Distribution of Gini indices calculated on bulk (ensemble average) in vivo DMS reactivities in 1 versus 2+ regions. f, Distribution of G+C content in 1 versus 2+ regions. g, Distribution of median percentage sequence conservation, calculated on a set of ten Gram-negative bacterial genomes, in 1 versus 2+ regions. For all box plots, boxes span the 25th to the 75th percentile. The center represents the median. Outliers (values below the 25th percentile − 1.5× the IQR or above the 75th percentile + 1.5× the IQR) are not shown. P values were calculated using a two-sided Wilcoxon rank-sum test.
