Extended Data Fig. 8: Transcriptome analysis of combinatorial targeting of AML differentiation regulators.

a) Single-cell gene expression heatmap showing the 50 most up-regulated (left) and down-regulated (right) genes upon GFI1 perturbation (Wilcoxon’s rank sum test) identified in ECCITE-seq data (Wang et al.). We compared cells expressing non-targeting control gRNAs (NT) and GFI1 targeting gRNAs for CaRPool-seq (top) and ECCITE-seq (bottom), respectively. The ECCITE-seq data was filtered using mixscape (Papalexi et al. 9) to remove unperturbed cells prior to identification of the most regulated genes. b) Gene module scores for the 50 most up-regulated and most down-regulated genes upon target gene perturbation identified in ECCITE-seq data as described in (a). The module scores show the average expression of perturbation-specific target genes per cell comparing cells with the indicated target gene perturbation to cells with non-targeting control gRNAs (one-tailed Kolmogorov–Smirnov test). c) EnrichR gene ontology analysis for biological processes (GOBP) for all single-gene perturbations (n = 28) using up to 100 significantly (p < 0.01, Wilcoxon’s rank sum test) upregulated genes per gene pair compared to NT condition. Shown are the -log10-transformed adjusted p-values for GO-terms with p < 0.00001 (Fisher’s exact test) in at least one condition. d-h) Comparison of transcriptional responses for double versus single perturbation. Heatmaps show deviation in average gene expression relative to unperturbed cells for the 20 most significantly regulated genes (Wilcoxon’s rank sum test). These heatmaps visualize a range of observed interactions between gene pairs, including cases where genes contribute equally to the dual perturbation response (d-f), and where one gene’s perturbation signature dominates over the other (g-h). Average heatmaps in g is accompanied by single-cell gene expression heatmap below.