Fig. 1

Schematic overview of the study design (a) and analysis pipeline (b). HSPCs were collected from human and mouse bone marrow, and single cell expression were profiled. Cells were clustered into number of groups and differential expressions (DEs) between pair-wised clusters were calculated. The correlation between pairs of genes was calculated with DEs for co-expression network reconstruction. CLR algorithm was used to correct background and build regulatory networks. c Curve of odds ratio for the enrichment of gene pairs in co-expression network in ChIP-chip versus all possible TF-TG pairs with different correlation cutoff. Higher co-expression between TF and target genes identified by ChIP-seq. Higher co-expressed pairs tended to appear more frequently in ChIP-chip validated pairs. d Venn diagram showing the overlap of co-expression gene pairs built with human and mouse datasets. Overlap of gene pairs between human and mouse is significantly higher than that expected from random selection (p < 1e-100)