Figure 2

Motif enrichments in enhancers are correlated with the expression of the TFs that bind to these motifs. a, b TFs’ expressions are correlated against the enrichment of their respective binding motifs in SE DHS sequences across the nine cell types. The master activator HNF1A (a) has a positive relationship between gene expression and motif enrichment, while the repressor GFI1 (b) has a negative relationship. The name of the motif, Spearman’s correlation (ρ), and sequence logo of the motif are displayed in the corner of the plots. c Boxplots of Spearman’s correlations of enrichments in four different regions for all database motifs against gene expression in the nine cell types. d Boxplots of Spearman’s correlations of motif enrichments between two different regions (listed in the strip titles at the top of the facets). For each motif and each pair of regions, we computed the enrichments of the motif in both sets of sequences and then computed the Spearman’s correlation between the two sets of enrichments. Within each facet, the boxplot of Spearman’s correlations (red, left) is also displayed alongside a boxplot of a null distribution (blue, right) generated by recalculating the Spearman’s correlations after shuffling cell assignments for one of the variables. P-values below the boxplots represent the significance of the distribution compared to the respective null distributions and are calculated with the Wilcoxon rank sum test. Gene expression is measured in reads per kilobase per million (RPKM). Motif sites were identified with FIMO, a tool for searching occurrences of known motifs in biological sequences [34]. Motif enrichment is calculated as the ratio of FIMO hits in the positive sequence set to FIMO hits in dinucleotide shuffled negative control (“Methods”).