1  What chromatin assays measure

Before you analyse data, you need to know what the data measure. Each assay in this book answers a different question about chromatin. This chapter explains what each assay measures, what it cannot see, and which controls it needs.

1.1 The question

What does each read in my experiment stand for? In ChIP-seq, a read marks a place where the antibody pulled down DNA. In ATAC-seq, a read marks a place where DNA was open enough for an enzyme to cut it. In bisulfite sequencing, a read tells you which cytosines were methylated. The analysis follows from this. The same software step can mean different things for different assays.

1.2 Chromatin in one page

DNA in the nucleus is wrapped around proteins called histones. One unit of DNA plus histones is a nucleosome. About 147 base pairs of DNA wrap around each nucleosome. Between nucleosomes there is free DNA, called linker DNA.

The tails of histone proteins carry chemical marks. A mark is named after the histone, the amino acid and the change. For example, H3K27ac means histone H3, lysine 27, acetylated. Different marks are found at different kinds of regions. The Roadmap Epigenomics project used five core marks and described them as follows (Roadmap Epigenomics Consortium et al. 2015):

Table 1.1: Five core histone marks and where they are found (Roadmap Epigenomics Consortium et al. 2015).
Mark Found at
H3K4me3 promoters
H3K4me1 enhancers
H3K36me3 transcribed regions
H3K27me3 regions repressed by Polycomb
H3K9me3 heterochromatin

A sixth mark, H3K27ac, separates active enhancers from poised ones (Creyghton et al. 2010). It is also found at active promoters.

Other proteins bind DNA directly. Transcription factors recognise short DNA sequences called motifs. They bind at precise sites and help switch genes on or off.

Some DNA is open: it is not covered by a nucleosome, so proteins can reach it. Promoters and active enhancers are usually open.

Finally, the DNA itself can be marked. In mammals, a methyl group is often added to a cytosine that is followed by a guanine. This pair is called a CpG. Methylation at a promoter is often linked to a silent gene.

1.3 Sharp and broad marks

Some histone marks cover a few hundred base pairs to a few kilobases around a site. Others cover large domains of tens of kilobases. ENCODE groups them as follows (ENCODE Project Consortium 2026):

  • Narrow (sharp) marks: H2A.Z, H3ac, H3K27ac, H3K4me2, H3K4me3, H3K9ac.
  • Broad marks: H3.3, H3K27me3, H3K36me3, H3K4me1, H3K79me2, H3K79me3, H3K9me1, H3K9me2, H4K20me1.
  • H3K9me3 is a special case, with its own depth requirement.

This difference changes the whole analysis. Broad marks need more reads (Chapter 6), a peak caller that looks for wide domains (Chapter 7), and a different way to compare conditions (Chapter 9).

NoteIn EpiNexus 1.2

You enter the target of each sample on the Samples page, for example “H3K27ac” or “CTCF”. EpiNexus decides from the target if the peaks are narrow or broad. It treats H3K27me3, H3K36me3, H3K9me3, H3K9me2, H3K79me2, H4K20me1, H3K4me1 and H2AK119ub1 as broad. All other targets, including transcription factors, are treated as narrow.

1.4 The assays

ChIP-seq

In ChIP-seq (chromatin immunoprecipitation followed by sequencing), cells are fixed so that proteins stick to DNA. The chromatin is then broken into small pieces, usually by sonication. An antibody pulls down the pieces that carry the target protein or histone mark. The DNA in these pieces is sequenced (Johnson et al. 2007; Barski et al. 2007).

ChIP-seq works for transcription factors and histone marks. It needs many cells, and its background is fairly high. So it needs more reads than the newer methods below.

CUT&RUN

In CUT&RUN, cells are not fixed. The antibody binds its target inside the nucleus. Then a protein called protein A–MNase binds the antibody. When it is switched on with calcium, MNase cuts the DNA around the target. The cut pieces leave the nucleus and are sequenced (Skene and Henikoff 2017). Only the pieces near the target are released, so the background is very low. The authors report that CUT&RUN needs “only ~1/10th the sequencing depth” of ChIP-seq (Skene and Henikoff 2017).

CUT&Tag

CUT&Tag works like CUT&RUN, but with a different enzyme. Protein A is joined to the Tn5 transposase. Tn5 cuts the DNA near the target and adds the sequencing adapters at the same time. This makes the library quick to prepare. CUT&Tag works with few cells, even single cells, and has very low background. About 3 million paired-end reads are enough for histone marks (Kaya-Okur et al. 2019).

ATAC-seq

ATAC-seq does not use an antibody. It measures open chromatin. The Tn5 transposase cuts open DNA and adds adapters, as in CUT&Tag. DNA wrapped around nucleosomes is cut much less. So reads gather in open regions such as promoters and enhancers. ATAC-seq works with 500 to 50,000 cells (Buenrostro et al. 2013).

ATAC-seq fragments have typical sizes. Short fragments come from open DNA between nucleosomes. Longer fragments span one or more nucleosomes. The sizes repeat about every 200 bp, because “many fragments are protected by integer multiples of nucleosomes” (Buenrostro et al. 2013). The Omni-ATAC protocol lowers the share of reads from mitochondria and works on frozen tissue (Corces et al. 2017).

Bisulfite sequencing

Bisulfite sequencing measures DNA methylation. Sodium bisulfite changes cytosine (C) into uracil, which is read as thymine (T). Methylated cytosine stays a cytosine (Frommer et al. 1992). After sequencing, each read is compared with the genome. A C in the read means the cytosine was methylated. A T means it was not.

There are two common types:

  • Whole-genome bisulfite sequencing (WGBS) reads the whole genome. It measures almost every CpG (Lister et al. 2009).
  • Reduced-representation bisulfite sequencing (RRBS) first cuts the DNA with the enzyme MspI, which cuts at the sequence CCGG. Then it keeps only short fragments. In the first genome-scale study, this was predicted to cover about 1 million CpGs of the mouse genome (4.8 % of all CpGs). About half of them were in CpG islands, including parts of 90 % of all CpG islands (Meissner et al. 2008; Gu et al. 2011). RRBS needs fewer reads than WGBS, but it does not see most of the genome.

1.5 Comparing the assays

Table 1.2: What each assay measures.
Assay Measures Background Usual control Peak caller in EpiNexus
ChIP-seq Where a protein or histone mark is Fairly high Input DNA MACS3
CUT&RUN Where a protein or histone mark is Very low IgG (optional) SEACR or GoPeaks
CUT&Tag Where a protein or histone mark is Very low IgG (optional) SEACR or GoPeaks
ATAC-seq Where chromatin is open Low None MACS3
WGBS Methylation at almost every CpG — — none (regions, not peaks)
RRBS Methylation at CpG-rich regions — — none (regions, not peaks)

1.6 Controls

Controls help you separate real signal from noise. There are three kinds.

Input DNA is used with ChIP-seq. It is chromatin from the same cells that went through the same steps, but without the antibody. Reads in the input show where DNA is easy to break or sequence, or where the genome has extra copies. These places would look like peaks without the input. ENCODE asks for an input for every ChIP-seq experiment, “with matching run type, read length, and replicate structure” (ENCODE Project Consortium 2026).

IgG is used with CUT&RUN and CUT&Tag. IgG is an antibody that binds nothing in particular. It shows where the enzyme cuts without a real target. These assays have very low background, so IgG libraries are often small.

Spike-ins solve a different problem. Most analysis methods assume that the total amount of signal is the same in all samples. This is not true when a treatment removes a mark everywhere, for example a drug that blocks the enzyme that writes the mark. Then you need a reference that does not change. A spike-in is chromatin or DNA from another species, added in the same amount to every sample (Orlando et al. 2014). Spike-ins are needed “where changes in absolute amounts of the total signal are suspected to occur between different experimental conditions” (Chen et al. 2016). Chapter 10 explains how to use them.

NoteIn EpiNexus 1.2

On the Samples page, you mark a control by setting its target to input or IgG. EpiNexus pairs each input with the ChIP sample that has the same condition and replicate number. A sample without such a control uses the control of its condition with the lowest replicate number. If all of the run’s controls come from one condition, they also serve the other conditions, matched by replicate number where possible (Chapter 15). Controls are never called for peaks: they are only used as background. Spike-ins are set in the run wizard, under Advanced Parameters.

1.7 Pitfalls

WarningPitfall: analysing a broad mark like a sharp one

H3K27me3 covers large domains. If you call narrow peaks on it, one domain is cut into many small peaks. Make sure the mark is entered correctly, so that the right peak mode is used.

WarningPitfall: reading ATAC-seq peaks as binding sites

An ATAC-seq peak shows that DNA is open. It does not tell you which protein binds there. Motif analysis (Chapter 12) can suggest candidates, but it is not proof.

WarningPitfall: expecting RRBS to cover the whole genome

RRBS sees CpG-rich regions well, but most of the genome is not covered. A region that is not in your RRBS data is not “unmethylated”. It is simply not measured.

1.8 Further reading

  • Roadmap Epigenomics Consortium et al. (2015): reference epigenomes for many human tissues, and what each mark shows.
  • Landt et al. (2012): the ENCODE guidelines for ChIP-seq, including controls.
  • Skene and Henikoff (2017) and Kaya-Okur et al. (2019): the original CUT&RUN and CUT&Tag papers.
  • Buenrostro et al. (2013) and Grandi et al. (2022): ATAC-seq and its current protocol.
  • Meissner et al. (2008) and Lister et al. (2009): RRBS and WGBS.