Skip to main content
Fig. 5 | BMC Genomics

Fig. 5

From: Computational modeling of chromatin accessibility identified important epigenomic regulators

Fig. 5

TF models identify cell line specific and shared specific TFs with strong effects on chromosome accessibility. A. Boxplots indicating the relative importance of each TFs binding signals in chromatin accessibility prediction in GM12878 (up panel) and HepG2 (bottom) cell lines. B. Barplots showing the AUC change of taking out each TF binding signal for chromatin accessibility prediction in GM12878 (up panel) and HepG2 (bottom) cell lines. C. Scatterplot showing the consistency of TFs relative importance between GM12878 and HepG2 cell lines. The circles indicated GM12878 specific TFs, HepG2 specific TFs and common TFs. D. Barplots showing the AUC of GM12878 specific TF model with GM12878 TFs binding signals for GM12878 chromatin accessibility prediction and GM12878 specific TF model with HepG2 TFs binding signals for HepG2 chromatin accessibility prediction (left panel). Barplots showing the AUC of HepG2 specific TF model with HepG2 TFs binding signals for HepG2 chromatin accessibility prediction and HepG2 specific TF model with GM12878 TFs binding signals for GM12878 chromatin accessibility prediction (right panel). E. Barplots showing the AUC of GM12878 common TF model with HepG2 TFs binding signals for HepG2 chromatin accessibility prediction and GM12878 specific TF model with HepG2 TFs binding signals for HepG2 chromatin accessibility prediction (left panel). Barplots showing the AUC of HepG2 common TF model with GM21878 TFs binding signals for GM12878 chromatin accessibility prediction and GM12878 common TF model with GM12878 TFs binding signals for GM12878 chromatin accessibility prediction (right panel). AUCs of GM12878 model in GM12878 TF signals and HepG2 model in HepG2 TF signals were calculated by taking the average AUC of 10-folds cross validation. AUCs of GM12878 model in HepG2 TF signals and HepG2 model in GM12878 TF signals were calculated by applying the model to a different test dataset

Back to article page