Fig. 5

Transcriptional kinetics inferred from scRNA-seq data show that DV genes have a high burst size and are regulated in burst size or frequency. a UMAP clustering of single-cell RNA-seq (scRNA-seq) from DV-relevant clusters in wild-type and Toll mutant 2.5–3.5h embryos [9] based on the expression of DV genes identified by PRO-seq. b Schematic of the two-state transcriptional model used for transcriptome-wide inference of burst kinetics from scRNA-seq [65]. c Boxplots showing the burst size and frequency (log₂) of DV genes classified by the tissue of expression in DV-relevant UMAP clusters from wild-type scRNA-seq. d Boxplots of the burst size and frequency of genes classified by the presence of de novo identified promoter motifs and compared to all DV genes. e Correlations between transcriptional kinetics and PRO-seq promoter read counts (log₂). The mRNA level (log₂ TPM) of genes is denoted. f Plots showing the transcriptional kinetics of individual DV genes (dpp, CG45263, SoxN, Meltrin, twi, and sna) across DV-relevant UMAP clusters. Error bars show the 95% confidence intervals. Genes with statistically significant increases in bursting kinetics in the cluster of expression relative to the OFF clusters are denoted. g Heatmap of the coefficient of determination (R²) between the tissue-specific enrichment of various genomic datasets at DV enhancers and promoters compared to burst frequency (BF) and size (BS) differences for enhancer-paired DV genes with significant changes in kinetics between DV clusters (n = 29). Comparisons with significant correlations are denoted. See Additional file 6: Table S7 for R² and P-values. h Boxplots showing the fold change (log₂) in transcriptional kinetics between the active tissue relative to the inactive tissues for genes regulated by proximal (≤ 700 bp from the TSS) and distal enhancers. Significant differences (Wilcoxon rank-sum test) are indicated by asterisks, *** = P < 0.001. i (Top) Boxplots of the inferred transcriptional kinetics for enhancer-paired DV genes partitioned into classes based on whether they have a significant kinetic change in burst frequency (n = 8), size (n = 16), or both (n = 5) between the active and inactive clusters (see also Additional file 1: Fig. S8h). (Bottom) For each class, the log₂ fold change (active/inactive tissues) is plotted. Significant differences (Wilcoxon rank-sum test) are indicated by asterisks, * = P < 0.05, ** = P < 0.01, *** = P < 0.001. j Heatmap showing the coefficient of determination (R²) between the chromatin state change at DV enhancers and promoters compared to the kinetics inferred for each class. Comparisons with significant positive and negative correlations are denoted. k Schematic model of tissue-specific DV gene transcriptional activation