Fig. 1

Profiling candidate silencers across 97 biosamples from diverse origins. A Classification performance (AUROCs) of TREDNet models for silencers and enhancers in analyzed biosamples. AUROCs exhibit a correlation with GC contents and repeat densities of training sequence sets. Each dot represents a set of enhancers or silencers. B Distribution of 97 biosamples across cell types. C Distance of candidate silencers and enhancers to their nearest TSSs. D Expression of genes proximal to candidate silencers and enhancers. Markers and their flanking lines represent the medians and standard deviations of gene expression levels. Blue and orange asterisks on the top represent the significantly low and high expression levels, respectively, compared to all genes (\(p<0.05\)). E MPRA scores of candidate silencers and enhancers in three biosamples. F Performance of the TREDNet model on MPRA silencers. G Densities of TF ChIP-seq peaks within candidate silencers and enhancers across biosamples. \(**:p<{10}^{-10}\)