Fig. 5: Identification of keratinocyte homeostasis transcription factors.
From: Disease-linked regulatory DNA variants and homeostatic transcription factors in epidermis
a Combined differentiation effect in perturb-seq based on altered pseudotime, including both sequencing technologies and all three time-points. Factors may have significant FDR values despite small effects if the Z scores average to near zero but one sample is highly significant. Red dots, known factors. b Correlation of differentiation effects between CRISPR-flow (as mean change to the KRT10 High/Low distribution) and perturb-seq (as changes to mean pseudotime), both as Z-scored effect sizes, identifies the most potent TFs as outliers and highlights potential novel TFs (green labels). c Guide enrichment in differentiated cells of the strongest TFs (by significance) in CRISPR-flow (calculated as in panel 3c) and perturb-seq (one-sided Mann–Whitney for altered pseudotime vs all other cells with one guide). Depletions from differentiated cells reversed the sign of the −log P-value for visualization of the direction of the effect. d Physical STRINGdb protein-protein interactions (confidence 0.5) among the homeostasis TFs link known and novel factors, such as ZNF217 to RCOR1. e Correlation of broken motif enrichment in MPRA (Motifbreakr motif assignments, Fisher’s exact test, two-sided) with averaged CRISPR-flow/perturb-seq effects (Z-scored effect size, taking the larger value between methods) on differentiation highlight AP-1 (red dots) and SP1.
