Fig. 4

GeneWalk analysis of nascent transcriptome response to BRD4 inhibition in T-ALL cells. a Schematic of the experimental design in Winter et al. [55]. NET-seq was performed on JQ1-treated MOLT4 cells (1 μM for 2 h, alongside DMSO controls, two biological replicates each). JQ1 targets BRD4 and other BET bromodomain family members, causing BRD4 to dissociate from chromatin [55]. b Volcano plot showing the results of a differential expression (DE) analysis comparing RNA Polymerase II gene coverage between JQ1 and DMSO control samples. DE genes (N = 2692), indicated in red, were used as an input to GeneWalk. All other genes are depicted in black. c All enriched Biological Process GO terms (five enriched terms, Fisher exact test, FDR = 0.05) in JQ1 condition, ranked by fold enrichment, obtained by GO enrichment analysis using PANTHER [1]. Red line indicates a fold enrichment value of 1, indicating the background. d The number of different unique GO annotations (y-axis) that are significant (p-adjust < 0.1) as a percentage of all unique GO annotation terms across all JQ1 DE genes present in the GWN. Average Kendall’s tau rank order correspondences of predictions from GeneWalk and alternative methods (x-axis) over previously identified transcriptional regulators that are part of the JQ1-context (Additional file 3) [55, 56] MYC, MYB, RUNX1, RUNX2, TAL1, SATB1, ERG, ETV6, and TCF12. Error bars indicate standard error on the mean. e Distribution of Kendall’s tau rank order correspondences of predictions from GeneWalk and seven tested alternative methods (Table 1) to the ground truth benchmark of the JQ1-context where all gene GO annotations pairs mentioned in [55,56,57] are jointly top-ranked and all other gene–GO annotations pairs are jointly bottom ranked. All methods are ordered by the median of their Kendall’s tau distribution, indicating their relative performances. Statistical differences between GeneWalk (INDRA or PC) and other methods are determined with the Wilcoxon signed-rank sum test. See the “Methods” for details. f Bar chart of the area under receiver operating characteristic (AUROC) performance metric for GeneWalk and alternative methods (Table 1) on the benchmark described in (e) when considered as a binary classification task: identifying gene-function pairs as relevant or not. g Scatter plot with DE genes as data points showing the GeneWalk fraction of relevant GO terms over total number of connected GO terms (min_f, minimum value between INDRA and PC GWNs) as a function of the number of gene connections in the GWN (N^gene, again minimal value between INDRA and PC). The circle size scales with the differential expression significance strength (−log₁₀(p-adjust)) and the color hue with min_f. Twenty genes were identified with min_f > 0.5 and N^gene > 30 (gray-shaded area, see Table 2 for complete list). h GeneWalk results for the transcriptional regulator RUNX1 under JQ1 treatment. Annotated biological process terms are rank-ordered by gene FDR adjusted p value. Error bars indicate 95% confidence intervals of gene p-adjust. FDR = 0.05 (dashed red line) is also shown. See Additional file 1 for full details. i As in (h) for transcriptional regulator MYB. j As in (h) for transcriptional regulator BRCA1. INDRA annotations are indicated by class: DNA damage and repair (green), chromatin, and post-translational modifications (dark blue), signaling pathways and cellular responses (light blue), transcription and gene expression (yellow), metabolism (purple), and other GO terms (gray)