Can ATAC-seq Alone Identify Functional Enhancers?

ATAC-seq tells you where DNA is accessible. Accessibility correlates with regulatory activity, but it also marks promoters, insulators, paused elements, silenced-but-open regions, and structural sites with no regulatory role. An ATAC peak is a candidate, and the field's own data show that only a minority of accessible elements show measurable regulatory activity when tested.

ATAC-seq alone cannot identify functional enhancers. It can efficiently narrow the search.

The row that matters most for downstream work is the last one: even a confirmed active enhancer has no target gene annotation until one is measured.

Accessibility without activity. Some elements are open in a cell type but poised rather than active, marked by H3K4me1 without H3K27ac, or accessible due to a structural role. Deleting them produces no expression change.

Promoters dominate the peak list. A large fraction of accessible regions sit at TSSs. Filtering by distance and by TSS overlap removes them, but the residual set still contains insulators and CTCF sites.

Wrong cell type, wrong state. An enhancer active in hepatocytes may be closed in fibroblasts. ATAC in an accessible-but-irrelevant cell line answers a question nobody asked.

Depth-dependent peak sets. Peak numbers scale with sequencing depth and vary with the caller; comparisons across datasets require matched processing.

Bulk averaging. In heterogeneous tissue, an element active in 10% of cells may be invisible in bulk ATAC and appears only in single-cell data.

A practical threshold for taking a candidate into a functional experiment:

1. Accessible in the relevant cell type (ATAC or DNase).
2. Active mark present — H3K27ac, or eRNA by GRO-seq/PRO-seq, distinguishing active from poised.
3. Not a promoter — no TSS overlap, no H3K4me3-dominant profile.
4. Not a known insulator — check CTCF occupancy, unless that is what you are studying.
5. Plausible target gene — from contact data, ABC-model prediction, or colocalization; labelled as a hypothesis.
6. A functional filter — MPRA or a small reporter assay to rank allele- or element-level activity.

Only after these does a perturbation experiment become worth its cost. Skipping step 6 is common and is the main reason screens return so few functional elements.

Two structural reasons explain why accessibility over-predicts function:

• • Redundancy. Many accessible elements are shadow enhancers or cluster constituents whose individual deletion is buffered, so they are functional yet individually dispensable. Our piece on enhancer redundancy covers the design implications.
• • Weak elements are common. The distribution of regulatory effect sizes is heavily skewed: a small number of elements carry most of the effect, while many accessible elements contribute marginal amounts that require sensitized assays to detect.

The consequence for planning: expect a substantial fraction of accessible candidates to show no individual effect, and design replication and controls accordingly rather than treating those results as failures.

The last row is the only one that establishes function. Everything above it exists to make that experiment affordable and interpretable.

EDITGENE builds the perturbation-ready models that follow candidate identification: CRISPRi/a lines for element-level necessity testing, reporter lines for quantitative element activity, and deletion panels for confirmation.

Tell us the candidate set and the cell type. Contact EDITGENE to request a project consultation or quote.

Peak sets depend on assay, depth, and caller, and accessibility-function correlation varies by cell type and locus. Statements here describe general limitations rather than a fixed fraction of functional elements.

Functional status should be established by perturbation with appropriate controls, not inferred from chromatin signal.

• • Buenrostro JD, et al. Transposition of native chromatin for fast and sensitive epigenomic profiling. Nature Methods. 2013;10:1213–1218.
• • Kwasnieski JC, et al. High-throughput functional testing of ENCODE segmentation predictions. Genome Research. 2014;24:1595–1602.
• • Fulco CP, et al. Systematic mapping of functional enhancer–promoter connections with CRISPR interference. Science. 2016;354:769–773.
• • ENCODE Project Consortium. An integrated encyclopedia of DNA elements in the human genome. Nature. 2012;489:57–74.