Enhancers are non-coding regulatory DNA sequences that increase transcription of nearby genes by serving as binding platforms for transcriptional activators and coactivators. They are typically marked by DNase I hypersensitivity and H3K27ac histone modification, and many are cell-type specific, reflecting the distinct gene expression programs of different tissues and developmental states.
This track displays 512,577 candidate enhancers predicted across 97 ENCODE biosamples by TREDNet, a two-phase deep learning model trained on DNA sequence features. The predictions span primary cell types, cell lines, and bulk tissue samples. Each merged region is annotated with the complete set of biosamples in which it was predicted as an enhancer, and colored by whether it is predominantly active in tissues, cell types, or cell lines.
Each item represents a merged window where at least one of the 97 biosamples carried a candidate enhancer prediction. Items are colored by the dominant sample category:
Tissue – region active predominantly in primary tissue samples
Cell type – region active predominantly in primary cell type samples
Cell line – region active predominantly in cell line samples
Mixed – no dominant sample category
Mousing over a region shows the contributing sample names when 10 or fewer samples are active, or a count summary (e.g., "37 samples: 16 tissues, 11 cell types, 10 cell lines") when more. Clicking an item opens the detail view with the full sample list.
TREDNet is a two-phase deep learning model for predicting enhancers and silencers from DNA sequence. The authors trained separate TREDNet models for each of 111 ENCODE biosamples using DNase-seq, H3K27ac ChIP-seq, and H3K27me3 ChIP-seq peaks. Enhancer training sequences were defined as DNase-seq peaks overlapping H3K27ac peaks (but not H3K27me3 peaks in the central 400 bp). Predictions used 1-kb sliding windows scored by the model, with an enhancer score cutoff set at a false positive rate of 0.1 (control-to-positive ratio of 9:1) on held-out chromosomes 7 and 8. Regions scoring above this threshold and not also predicted as silencers were retained as candidate enhancers. 97 of the 111 biosamples yielded more than 5,000 candidate enhancers and were included in the final dataset. The TREDNet models achieved an enhancer area under the ROC curve (AUROC) of 0.96 on held-out test data. For further details see Huang and Ovcharenko (2024).
The per-biosample enhancer BED files were downloaded from Zenodo record 12523205 (file SilencerEnhancer.tar.gz). Biosample filenames used Cell Ontology (CL), Uberon (UBERON), Experimental Factor Ontology (EFO), and ENCODE New Term Request (NTR) identifiers, which were resolved to readable names using the EBI Ontology Lookup Service and the ENCODE project API. Overlapping per-sample predictions were merged with bedtools merge and annotated with the full list of contributing biosamples. The processing scripts are available at github.com/ucscGenomeBrowser/kent/.../scripts/tredNet and the build instructions are documented in src/hg/makeDb/doc/hg38/crPred.txt.
The data can be explored interactively in table format with the Table Browser or the Data Integrator and exported from there to spreadsheet or tab-separated tables. From scripts, the data can be accessed through our API, track=tredNetEnhancer.
For automated download and analysis, the annotation is stored in a bigBed file that can be downloaded from our download server. The file for this track is called tredNetEnhancer.bb. Individual regions or the whole genome annotation can be obtained using our tool bigBedToBed, which can be compiled from the source code or downloaded as a precompiled binary for your system. Instructions for downloading source code and binaries can be found here. The tool can also be used to obtain features within a given range, e.g., bigBedToBed http://hgdownload.soe.ucsc.edu/gbdb/hg38/tredNet/tredNetEnhancer.bb -chrom=chr21 -start=0 -end=100000000 stdout.
The original annotation source data can be downloaded from Zenodo record 12523205.
Huang D, Ovcharenko I. The contribution of silencer variants to human diseases. Genome Biol. 2024 Jul 8;25(1):184. PMID: 38978133; PMC: PMC11232194