Figure 5—figure supplement 2. | Epigenomic landscapes of retinal rods and cones

Open accessCopyright infoDownload PDFDownload figures

Epigenomic landscapes of retinal rods and cones

Figure 5—figure supplement 2.

Affiliation details

Johns Hopkins University School of Medicine, United States; The Salk Institute for Biological Studies, United States; Howard Hughes Medical Institute, The Salk Institute for Biological Studies, United States; Janelia Research Campus, Howard Hughes Medical Institute, United States; University of California San Diego, United States; The University of Western Australia, Australia; Johns Hopkins University, United States
Figure 5—figure supplement 2.
Download figureOpen in new tabFigure 5—figure supplement 2. DNA regulatory sequences inferred from retinal chromatin accessibility yield gkm-SVM scores which predict enhancer activity in a massively parallel reporter assay.

(A) A scatterplot showing that retinal expression levels of >3000 candidate retina, brain, heart, and liver CREs (Shen et al., 2016) are strongly correlated with the number of replicates in which each candidate CRE barcode was detected in the RNA sample. In addition, candidate CREs with high expression have higher scores using the regulatory vocabulary trained on the WT retina ATAC-seq dataset. Error bars show mean +/- 1 S.D. for each of the four sets of data points. (B) Relative to all candidate CREs, the top-scoring 10% of 36,005 candidate CRE constructs are strongly enriched in highly expressed sequences (i.e., those detected in all three replicates) when gkm-SVM is trained on retinal ATAC-seq peaks or DHS. There is no enrichment when training is performed with chromatin features from non-retinal cell types: p300-bound enhancers in melanocytes (Gorkin et al., 2012), GATA1-bound enhancers in megakaryocytes (Pimkin et al., 2014), and DHSs in lymphoblasts (Lee et al., 2015). (C) For retinal ATAC-seq or DHS, the mean gkm-SVM score is higher for sequences detected in all three replicates, than for sequences detected in zero, one, or two replicates, or when gkm-SVM is trained on chromatin features from non-retinal cell types.

DOI: http://dx.doi.org/10.7554/eLife.11613.017