Analysis of functional importance of binding sites in the Drosophila gap gene network model

Abstract

Background: The statistical thermodynamics based approach provides a promising framework for construction of the genotype-phenotype map in many biological systems. Among important aspects of a good model connecting the DNA sequence information with that of a molecular phenotype (gene expression) is the selection of regulatory interactions and relevant transcription factor bindings sites. As the model may predict different levels of the functional importance of specific binding sites in different genomic and regulatory contexts, it is essential to formulate and study such models under different modeling assumptions.

Results: We elaborate a two-layer model for the Drosophila gap gene network and include in the model a combined set of transcription factor binding sites and concentration dependent regulatory interaction between gap genes hunchback and Kruppel. We show that the new variants of the model are more consistent in terms of gene expression predictions for various genetic constructs in comparison to previous work. We quantify the functional importance of binding sites by calculating their impact on gene expression in the model and calculate how these impacts correlate across all sites under different modeling assumptions.

Conclusions: The assumption about the dual interaction between hb and Kr leads to the most consistent modeling results, but, on the other hand, may obscure existence of indirect interactions between binding sites in regulatory regions of distinct genes. The analysis confirms the previously formulated regulation concept of many weak binding sites working in concert. The model predicts a more or less uniform distribution of functionally important binding sites over the sets of experimentally characterized regulatory modules and other open chromatin domains.

Figure 1

The output for Models 2, 3 and 4 as compared to protein and mRNA concentration profiles from the FlyEx and SuperFly databases. Upper and lower rows of panels for each model present results for late (T7) cleavage cycle 14A and for mRNA and protein respectively. Though there are some defects in predicted patterns, all the models correctly reproduce the main features of the system.

Figure 2

Plot of the TFBS regulatory weights estimated with the RSS measure and in frame of model 4 relative to site position in a regulatory region. The binding sites for different TF are shown in different color. The transcription start site is at zero position. Results for hb regulatory region are presented relative to TSS of the longest transcript. Sites within CRM are shown as triangles, sites in the DNase I accessible region are marked with circles and rombs present sites in both regions. The empty triangles denote the sites annotated with DNase I footprinting.

Figure 3

The histograms of regulatory weights calculted with RSS and wPGP measures. The thresholds are clearly seen - w^rss = 0.05 and w^wpgp = 0.02

Figure 4

Correlation matrix between spatio-temporal distributions of TFBS impacts (model 3). The diffusion rate parameter was set to zero during calculation. The color in the figure reflects the correlation strength between impact distributions for each pair of TFBSs. The sites are ordered alphabetically - first by target gene (gt, hb, kni, and Kr) and then by TF (Bcd, Cad, Gt, Hb, Hkb, Kni, Kr, and Tll) in each group. The clusters of highly correlated sites appear as rectangles of yellow or red color. Arrow with asterisk marks the matrix region which shows high positive correlation between Bcd sites in the gt regulatory region and Hb sites in the hb regulatory region.

Figure 5

Spatial distribution of impact on gap gene expression patterns of each TFBS in the hb regulatory region at temporal class 8 (model 4). The sites are ordered according to their coordinate. Sites from different parts of the regulatory region modulate expression in different spatial locations. Some visually identifiable clusters of functionally important sites correspond to CRMs, e.g. anteriory expressed CRMs hb_HB747, hb_0.7 and hb_0.8 include sites that cluster at the bottom of this picture.

Figure 6

Spatial distribution of impact on gap gene expression patterns of each TFBS in the hb regulatory region for T8 (model 4). The sites are ordered according to the TF and then by coordinate. Different sites of the same TF may have different spatial effects in the model. The insert shows that the strength of site impact can vary between several adjacent nuclei due to regulatory interactions.