Skip to main content
Fig. 3 | Genome Biology

Fig. 3

From: An explainable artificial intelligence approach for decoding the enhancer histone modifications code and identification of novel enhancers in Drosophila

Fig. 3

Putative enhancers make 3D contacts with expressed genes. A We split enhancers into the following: (i) proximal if they are located within 5 Kb of a promoter, (ii) distal if they are further than 5Kb from any promoters and make 3D contact with promoters, (iii) proximal only if they do not have enriched Hi-C contacts more than 5 kb away but were within 5 Kb of a promoter, and (iv) neither if they are further than 5Kb from any promoter and do not make 3D contacts with any promoters. Top: Putative, common, and STARR-seq enhancers have enriched 3D contacts with regions containing proximal (within 5 Kb from enhancer) or a distal (further than 5 Kb from the enhancer) promoters. We considered the case of BG3 and S2 cells respectively. Bottom: log2(observed/expected) based on whole genome distribution of the different annotations. B The size of distal only and proximal putative enhancers in BG3 and S2 cells on log2 scale. There is negligible difference between distal only or proximal putative enhancers (Mann-Whitney U test of log2 of size; p value < 1.34 × 10−5 for BG3 and p value = 0.09 for S2). C Majority of the enhancers that make 3D contacts with genes contact expressed genes, but there are significantly more distal only than proximal enhancers that contact expressed genes (Fisher’s extract test; p value: n.s. ≥ 0.05, *p value < 0.05, ** < 0.01 and *** < 0.001). D Top: Expression (FPKM) for proximal and distal only putative enhancers on log2 scale. We considered the maximum expression, in the case where promoters of multiple genes were contacted. There is a higher expression for genes controlled by distal only enhancers compared to proximal ones (Mann-Whitney U test of log2 of FPKM; p value < 2.2 × 10−16 for BG3 and S2). Bottom: Expression (FPKM) for proximal and distal only background regions on log2 scale. There is a higher expression for genes contacted by distal only background regions compared to proximal ones (Mann-Whitney U test of log2 of FPKM; p value < 2.2 × 10−16 for BG3 and S2). In BG3 cells, distal only enhancers have a mean log2 of FPKM of 6.08, while distal background regions of 5.82 (p value < 2.2 × 10−16). Similarly, in BG3 cells, proximal enhancers have a mean log2 of FPKM of 4.59 and proximal background regions of 3.77 (p value < 2.2 × 10−16). In S2 cells, distal only enhancers have a mean log2 of FPKM of 6.54, while distal background regions of 5.94 (p value < 2.2 × 10−16). Similarly, in S2 cells, proximal enhancers have a mean log2 of FPKM of 4.86 and proximal background regions of 3.81 (p value < 2.2 × 10−16). Note that in each case, we performed a Mann-Whitney U test

Back to article page