Fig. 5

Prediction of gene expression patterns from combinatorial binding of NF-κB subunits using logistic regression analyses. a Left: Regulation by non-canonical NF-κB versus no regulation. Right: Regulation by canonical NF-κB versus no regulation. Note that both, up- and downregulated genes are included in the categories ‘canonical’ regulated or ‘non-canonical’ regulated genes. Different parameter choices were systematically analyzed to evaluate the performance of the prediction framework for the logistic regression models (see Additional file 2: Figure S5). The y-axis shows the combinations of NF-κB subunits, which are listed as binary vectors (1 = presence, 0 = absence). The x-axis shows parameter estimates of the logistic regression models, which can be interpreted as log odds ratios and are indicated by circles. Vertical lines show confidence intervals (95 %). Filled circles indicate that parameters are significantly different from zero (P <0.01). Missing estimates and confidence intervals occur when the specific subunit combination was not observed in the dataset. Bar plots on the right side indicate the percentage of bootstrap samples in which each parameter was significantly different from zero (P <0.01). The dashed line marks 90 %. See Additional file 1: Supplemental Experimental Procedures for details. b Schematic presentation of the dependency of gene activation or repression on the distance and subunit occupancies of NF-κB binding regions