Fig. 3: Overview of SVM model and candidate silencer element predictions.
From: Candidate silencer elements for the human and mouse genomes
a Schematic overview of the gkmSVM model development for candidate silencer predictions. The trained gkmSVM model performance on test data shown by receiver operating characteristic (ROC) curve (b) and precision recall curve (c). d Heatmap of enrichment (-log(P-value)) of TF motifs at SVM positively predicted uncharacterized CREs (candidate silencers) relative to SVM negatively predicted uncharacterized CREs across 52 cell types and tissues from Roadmap. Barplots presenting the count of predicted candidate silencer elements across different cell types and tissues in human from Roadmap (e) and mouse genomes from ENCODE (g). Distribution of cell-type-specific candidate silencer elements in human (f) and mouse genomes (h). Pie chart showing the genomic distributions of candidate silencer elements in human (i) and mouse (j).
