Figure 3: TF combinations drive modular gene regulatory networks.

A. Schematic to estimate gene regulatory networks driven by perturbation clusters. Cells in each perturbation cluster are grouped, followed by differential expression analysis.

B. For each perturbation cluster, shown is the number of differentially expressed genes induced and repressed. (left) The number of unique TFs in the perturbation cluster. (right) The number of unique TF combinations.

C. Across all differentially expressed genes identified in B, shown is the specificity across modules (1 - fraction of occurrence across modules). A specificity score of 0 indicates a gene is found in all modules.

D. Expression of Mylpf in cells across the 51 perturbation clusters.

E. Gene ontology enrichment for high specificity and low specificity genes.

F. The heatmap indicates the overlap fraction of TFs across perturbation clusters. The histogram quantifies Jaccard similarity of TFs across perturbation clusters.

G. Heatmap indicating gene expression patterns (bottom) across perturbation clusters. Each column indicates a distinct TF combination (middle), grouped by batch/assay and perturbation cluster (top).

H. The heatmap illustrates the expression of 869 differentially expressed genes in cells induced with Gata6 alone or Gata6 + TF_X compared to control MEFs. TF_X spans 11 different TFs from different TF families (bold). For each Gata6 + TF_X pair, the values on the right column indicate Euclidean distance relative to control cells.