E. coli gene regulatory networks are inconsistent with gene expression data

Abstract

Gene regulatory networks (GRNs) and gene expression data form a core element of systems biology-based phenotyping. Changes in the expression of transcription factors are commonly believed to have a causal effect on the expression of their targets. Here we evaluated in the best researched model organism, Escherichia coli, the consistency between a GRN and a large gene expression compendium. Surprisingly, a modest correlation was observed between the expression of transcription factors and their targets and, most noteworthy, both activating and repressing interactions were associated with positive correlation. When evaluated using a sign consistency model we found the regulatory network was not more consistent with measured expression than random network models. We conclude that, at least in E. coli, one cannot expect a causal relationship between the expression of transcription and factors their targets, and that the current static GRN does not adequately explain transcriptional regulation. The implications of this are profound as they question what we consider established knowledge of the systemic biology of cells and point to methodological limitations with respect to single omics analysis, static networks and temporality.

Figure 1.

Example of consistency vectors for a gene g regulated by two transcription factors t₁ and t₂. Labels for three different contrasts are shown in brackets next to the vertices.

Figure 2.

Distribution of Pearson correlation coefficients for TF and target gene/unit pairs. (A, D) Comparison between correlation of all possible TF-target pairs and all known interactions. (B, E) Comparison between correlation of known activations and repressions. (C, F) Comparison between correlation of known activations and repressions where the TF is the only regulator of the target. Dashed vertical lines indicate mean correlation for each set of interactions.

Figure 3.

Evaluation of inconsistency load in regulatory network and perturbed network models. (A, D) Global inconsistency load in regulatory networks compared to two random networks models. For the random models, each experiment was repeated 200 times. (B, E) Distribution of edge inconsistency for repressing and activating interactions. (C, F) Distribution of edge inconsistency for repressing and activating interactions targeting genes/TUs with only one regulator. Dashed vertical lines in (B, C, E, F) indicate mean inconsistency for each set of interactions.

Figure 4.

Evaluation of inconsistency load of E. coli across contrasts and experimental evidence types. (A) Distribution of inconsistency load across the 655 contrasts. (B) Comparison between inconsistency load in contrasts with and without perturbation (e.g. drugs and experimental contitions). (C) Comparison between number of up- or downregulated genes in sign consistency model for in contrasts with and without perturbation. (D) Relationship between number of up- or downregulated genes and inconsistency load in contrasts. (E) Comparison between inconsistency of interactions with strong and weak experimental evidence. (F) Comparison between different common experimental evidence types for regulatory interactions. Evidence types: binding of cellular extracts (BCE), site mutation (SM), binding of purified proteins (BPP), gene expression analysis (GEA), human inference based on similarity to consensus sequences (HIBSCS), automated inference based on similarity to consensus sequences (AIBSCS). Significance in (B, C, E) was computed using a Mann–Whitney U-test.