Fig. 6
From: Genome-wide prediction of DNase I hypersensitivity using gene expression
Predicting regulome using PDDB.a–c Predicted DH in promoter regions of FBL (a), LIN28A (b), and BLMH (c) in P493-6 B cell lymphoma and H9 embryonic stem cells. For H9, PDDB contains multiple replicate samples, which produced similar results. One replicate is shown here and the other replicates are shown in Supplementary Fig. 31. d Predicted DH at 6931 SOX2 binding sites in 2000 PDDB samples. Each column is a sample, and each row is a binding site. Values within each row are standardized to have zero mean and unit SD before visualization. e Relative DH enrichment level when comparing SOX2 binding sites with random sites (Supplementary Methods). f Predicted DH at SOX2 binding sites in H7 stem cells after 2, 5, and 9 days of differentiation. True DH from undifferentiated H7 cells and cells at differentiating day 14 in the training data are also shown. Rows are SOX2 sites and columns are time points. Values within each row are standardized before visualization. g Predicted DH at SOX2 binding sites are compared with predicted DH at 10,000 random DHSs. At each time point, DH values from all sites are displayed using a boxplot, which shows the median (central line), interquartile range (IQR, the 1st (Q1) to 3rd (Q3) quartiles, box), and 1.5 × IQR from the Q1 and Q3 (lower and upper whiskers), respectively. h Predicted DH at 2011 MEF2A binding sites in 1061 MEF2A-expressing PDDB samples (Supplementary Methods). Each column is a sample. Each row is a MEF2A binding site. Values within each row are standardized before visualization. Samples and DHSs were clustered. i The highlighted region in h that shows DHS-clusters with increased DH in muscle, lymphoblastoid, and brain-related samples, respectively
