ChIP Sequencing · Chromatin Biology

Map protein–DNA interactions. Reveal chromatin state.

End-to-end Chromatin Immunoprecipitation Sequencing (ChIP-Seq) analysis for genome-wide mapping of transcription factor binding sites and histone modification patterns — from raw reads to annotated peaks and motif enrichment.

Plan Your Project → See the Workflow
98.6%
Mean primary alignment rate
~18K
High-confidence peaks (TF example)
2
Peak modes: narrow & broad
FRiP
Enrichment metrics throughout
About the Service

Genome-wide mapping of protein–DNA interactions

Nucleome Informatics offers end-to-end ChIP-Seq data analysis services for genome-wide mapping of transcription factor binding sites and histone modification patterns.

ChIP-Seq combines chromatin immunoprecipitation with high-throughput sequencing to identify regions of the genome bound by a protein of interest, providing insight into gene regulation and chromatin state.

Our analysis combines sequencing quality control, reference alignment, duplicate and quality filtering, peak calling suited to narrow (transcription factor) or broad (histone mark) binding profiles, peak annotation and motif enrichment analysis.

  • Sequencing data quality assessment and preprocessing
  • Reference genome alignment and read-level processing
  • Duplicate removal and quality filtering
  • Peak calling for narrow and broad profiles
  • Peak annotation relative to genes and genomic features
  • Motif enrichment analysis
  • Differential binding analysis across conditions or replicates
  • Secure delivery of raw, intermediate and final analysis files

Narrow & Broad Peak Support

Transcription factor binding produces sharp, narrow peaks. Histone modifications often span broader domains. Our pipeline selects the appropriate calling strategy for your target.


Quality-Controlled Throughout

Library complexity, cross-correlation and fraction-of-reads-in-peaks (FRiP) metrics are tracked at every stage — so you know your enrichment is real.

Technology Highlights

Built for chromatin biology

Flexible peak calling, comprehensive annotation and reproducible reporting — designed around the questions chromatin researchers actually ask.

Flexible Peak Calling

Narrow-peak calling for transcription factors and broad-peak calling for histone modifications, using established approaches.

Reference Genome Support

Analysis against GRCh38/hg38 or a client-specified reference build, as appropriate to your study.

Peak Annotation

Annotation of peaks relative to nearest genes and genomic features — promoter, exon, intron, intergenic.

Motif Enrichment

De novo and known transcription-factor motif enrichment analysis within called peaks.

Quality-Controlled Analysis

Library complexity, cross-correlation and FRiP metrics tracked throughout the workflow.

Reproducible Reporting

Structured peak sets, QC summaries and project reports delivered in a consistent format.

Learning Path

From first principles to frontier methods

New to chromatin biology? Start with Foundations. Already running experiments? Jump to Practitioner or Expert. Each level builds on the one before it, so you can read the page as a short course or dip in where you need to.

The big idea in plain language: ChIP-Seq tells you where on the genome a protein, or a chemical mark on histones, is sitting. That location map is the first step to understanding which genes a cell can switch on.

LEVEL 1

DNA is packaged, not naked

Each human cell fits roughly two metres of DNA into a nucleus by wrapping it around histone proteins. This DNA–protein complex is chromatin, and how it is wrapped helps decide which genes can be read.

LEVEL 1

Transcription factors are the readers

Transcription factors (TFs) recognise short DNA sequences and turn nearby genes up or down. Knowing where a TF binds shows which genes it can regulate.

LEVEL 1

Histone marks are the annotations

Chemical tags on histones, such as H3K4me3 at active promoters or H3K27me3 in Polycomb-silenced regions, label the functional state of each stretch of genome.

LEVEL 1

ChIP-Seq is a molecular snapshot

Proteins are glued to DNA, the DNA is cut into pieces, an antibody pulls out the protein you care about, and the DNA it was holding is sequenced. Think of it as a molecular highlighter for the genome.

LEVEL 1

A peak is a pile-up of reads

Sequenced fragments are mapped back to the genome. Where many fragments stack up, the protein or mark was probably present. That stack is called a peak.

LEVEL 1

The input control is your blank

Some regions look sticky or open for purely technical reasons. A matched input sample, without the antibody pull-down, shows that background so it can be separated from true signal.

For scientists planning or interpreting a project: how experiment design, depth and quality metrics translate into trustworthy peak sets.

LEVEL 2

Narrow vs broad targets

TFs give tight peaks a few hundred bp wide, while many histone marks form kilobase-scale domains, so peak callers run in different modes. ENCODE4 asks for at least 10 M usable fragments per replicate for TFs (more than 20 M recommended); broad marks need at least 20 M (more than 45 M recommended).

LEVEL 2

Replicates and controls

ENCODE4 expects two or more biological replicates, and every ChIP needs a matched input control with the same run type and read length. Replicates are what let you separate biology from noise.

LEVEL 2

The QC ladder

Q30 and adapters, then mapping rate, library complexity (NRF, PBC1, PBC2), strand cross-correlation (NSC, RSC) and FRiP. Each rung catches a different failure: poor sequencing, wrong reference, over-amplification or weak enrichment.

LEVEL 2

Peak-calling statistics

Callers model local background against input with a Poisson-type test, correct for multiple testing (FDR or q-value) and report summit position, fold enrichment and significance for each peak.

LEVEL 2

Annotation and motifs

Peaks are assigned to nearest genes and features (promoter, exon, intron, intergenic). Motif analysis then asks whether the TF's own recognition sequence is enriched near summits, a built-in sanity check on the antibody.

LEVEL 2

Differential binding

To compare conditions, replicate-aware count models work on a consensus peak set, normalise library sizes and test log2 fold changes with FDR control.

For computational biologists and reviewers: reproducibility frameworks, artefacts, quantitation and the machine-learning models now built on ChIP-seq-type data.

LEVEL 3

Reproducibility with IDR

The Irreproducible Discovery Rate ranks peaks across replicates and keeps those with consistent rank. ENCODE4 also uses pseudoreplicates: the rescue and self-consistency ratios should both stay below 2.

LEVEL 3

Artefact control

Blacklisted regions, hyper-ChIPable high-occupancy loci and antibody cross-reactivity can create convincing but false peaks. Blacklist filtering, motif checks and orthogonal controls protect your conclusions.

LEVEL 3

Absolute quantification

Standard ChIP-seq is relative. Adding spike-in chromatin from another species (ChIP-Rx or quantitative ChIP-seq) gives a reference that can reveal global gains or losses of a mark between conditions.

LEVEL 3

Base-pair resolution

ChIP-exo and ChIP-nexus trim DNA with exonuclease to expose protein–DNA footprints, and BPNet-style neural networks trained on such data describe motif syntax at single-base resolution.

LEVEL 3

Multi-omic integration

Combining several marks into chromatin states, then layering accessibility, 3D contacts and expression, turns peak lists into regulatory maps of promoters, enhancers and silenced domains.

LEVEL 3

Sequence-to-function AI

Models such as BPNet, ChromBPNet, Enformer and AlphaGenome learn from epigenomic tracks to predict how non-coding variants change regulatory activity, helping prioritise variants for follow-up.

Peak
A genomic region with more reads in the ChIP than expected from the input.
Summit
The single position of highest signal inside a peak; used to centre motif searches.
Input / control
Matched sample processed without antibody enrichment; models background.
FRiP
Fraction of Reads in Peaks, a signal-to-noise summary of the immunoprecipitation.
Fragment length
Size of the DNA pieces captured; estimated from the shift between forward and reverse reads.
Pseudoreplicate
A random half of the reads from a sample, used to test self-consistency.
Motif
A short, recurring DNA sequence pattern that a TF prefers to bind.
Blacklist region
Genomic locus known to give artefactual signal; excluded from analysis.
The Science

How a protein's address becomes a peak

ChIP-Seq turns a snapshot of where proteins sit on DNA into sequencing reads, and statistics turn those reads into binding sites.

Figure 1 From living chromatin to sequencing reads schematic
1 Fix2 Shear3 Pull down4 Sequence5 Analyse TF formaldehyde TF sonication bead TF washedaway library fragment short reads ChIP Input Peak Formaldehyde locksprotein onto DNA Chromatin cut into~200–600 bp pieces Antibody + beads capturetarget-bound fragments Reverse crosslinks, buildlibrary, read the ends Reads pile into peaks(Nucleome analysis)
Figure 1 | The ChIP-Seq principle. (1) Protein–DNA contacts are fixed with formaldehyde. (2) Chromatin is fragmented, typically by sonication or enzymatic digestion. (3) An antibody coupled to beads captures fragments carrying the target protein or histone mark, while unbound chromatin is washed away. (4) Crosslinks are reversed, DNA is purified, and a sequencing library is prepared and read. (5) Reads are aligned to the genome; fragments accumulate over bound regions to form peaks relative to the matched input control.
Analysis Workflow

From raw reads to annotated peaks

Each sample passes through a quality-controlled analysis workflow designed to move from raw sequencing data to annotated, high-confidence binding-site datasets.

Figure 2 QC-gated analysis pipeline, file formats and decision points infographic
STAGE 01

Raw read QC

Per-base quality, adapter content, read-length distribution; trimming where needed.

FASTQ
Q30 · adapters
STAGE 02

Reference alignment

Reads aligned to GRCh38/hg38 or a client-specified build; primary alignment and mapping quality recorded.

BAMBAI
alignment rate · MAPQ
STAGE 03

Filtering

PCR duplicates, low-quality and multi-mapped reads removed; problematic regions can be masked.

filtered BAM
NRF · PBC1 · PBC2
STAGE 04

Fragment modelling

Strand cross-correlation estimates fragment length and separates enrichment from background.

cross-correlation
NSC · RSC
STAGE 05

Peak calling

Narrow or broad mode against input control, with significance and FDR control.

narrowPeakbroadPeakbigWig
FRiP · peak count
STAGE 06

Replicate concordance

Peaks compared across replicates and pseudoreplicates to retain reproducible sites.

consensus BED
IDR ratios < 2
STAGE 07

Annotation & motifs

Nearest-gene and feature annotation, plus known and de novo motif enrichment.

annotated TSVmotif report
motif specificity
STAGE 08

Differential & report

Optional differential binding across conditions, then QC and project reports delivered securely.

reporttables
FDR-controlled
Figure 2 | Decision points. Amber gates mark where metrics are reviewed before a sample proceeds. Gate metrics are defined in the QC section below; reference values from the ENCODE4 data standards are given there for context.
01

Data QC & Pre-processing

Sequencing quality, read quality and adapter content assessed; quality filtering applied.

02

Reference Alignment

High-quality reads aligned to the selected reference genome for downstream peak analysis.

03

Duplicate Removal

PCR duplicates and low-quality or multi-mapped reads removed for reliable peak calling.

04

Peak Calling

Enriched regions identified relative to input/control — narrow or broad as appropriate.

05

Peak Annotation — Called peaks annotated relative to nearest genes and genomic features.

06

Motif Enrichment — Peak sequences analysed for known and de novo sequence motifs.

07

Reporting & Delivery — Peak statistics, annotation summaries and QC metrics delivered securely.

Differential Binding Analysis

Where multiple conditions or biological replicates are available, differential binding analysis can be performed on the same processed data to identify condition-specific or treatment-responsive binding regions.

Qc In Depth

Reading the diagnostic plots like a reviewer

Good ChIP-seq papers show their quality control. These are the plots reviewers look for, with the community reference values from the ENCODE consortium so you can benchmark any dataset.

Figure 3 Community reference standards for ChIP-seq data source: ENCODE4
Transcription factor
min 10 M · recommended >20 M
Narrow histone marks
min 10 M · recommended >20 M
Broad histone marks
min 20 M · recommended >45 M
010 M20 M30 M40 M50 M
Replication≥ 2biological replicates, each with a matched input control of the same run type and read length.
Library complexityNRF > 0.9PBC1 > 0.9 and PBC2 > 10 indicate low redundancy and good usable depth.
ReproducibilityIDR < 2Rescue and self-consistency ratios both below 2; IDR cut-off 0.05 for replicated peaks.
Figure 3 | Depth and quality benchmarks. Bars show minimum (dark) and recommended (light) usable fragments per replicate. Values are taken from the ENCODE4 transcription-factor and histone ChIP-seq data standards. For H3K9me3 in tissues and primary cells ENCODE4 specifies a higher total mapped-read depth because of repetitive regions. FRiP is reported for comparing similar experiments rather than as a pass/fail threshold, and project-specific depth is finalised with your team according to target and study design.
Figure 4 Strand cross-correlation simulated
Figure 4 | Cross-correlation separates enrichment from noise. Forward- and reverse-strand reads are shifted against one another. A successful immunoprecipitation shows a strong peak at the fragment length (~185 bp here); a failed one is dominated by the artefactual phantom peak at read length. NSC and RSC summarise the contrast.
Figure 5 Fingerprint (Lorenz) plot simulated
Figure 5 | Enrichment fingerprint. Genomic bins are ranked by coverage. Input follows the diagonal, because reads are spread almost evenly. A good ChIP concentrates most reads into a small fraction of bins, bending the curve far below the diagonal; weak enrichment stays close to input.
Table 1 Metrics at a glance
FRiPreads in peaks ÷ total usable reads

Signal-to-noise of the pull-down; compare between similar targets and antibodies.

NRFdistinct mapped positions ÷ total mapped reads

Fraction of non-redundant reads; low values mean over-amplification.

PBC1 · PBC2n₁ ÷ n≥1 · n₁ ÷ n₂

n₁, n₂ = positions with exactly one or two reads; n≥1 = positions with at least one.

NSC · RSCNSC = CCmax ÷ CCmin · RSC = (CCfrag − CCmin) ÷ (CCphantom − CCmin)

Normalised and relative strand cross-correlation from Figure 4.

Peak significancep = P(X ≥ k | λlocal), λlocal = max(λbg, λ1k, λ5k, λ10k)

Poisson test against a locally estimated background, as used by MACS-family callers; q-values apply Benjamini–Hochberg correction.

IDR ratiosrescue = max(Np,Nt) ÷ min(Np,Nt) · self = max(N1,N2) ÷ min(N1,N2)

Peak counts from true replicates (Nt), pooled pseudoreplicates (Np) and self-consistency pairs (N1, N2).

Peak-Level Analytics

What the peaks actually look like

Aggregate profiles, signal heatmaps and reproducibility plots are the standard way to show that peaks are real, well-centred and consistent across replicates.

Figure 6 Signal profiles and heatmaps around peaks simulated
Figure 6 | Aggregate profiles and ranked heatmaps. (a) A transcription factor gives a sharp, summit-centred enrichment (±2 kb window; each row is one peak, ranked by signal). (b) A promoter-associated histone mark such as H3K4me3 gives a bimodal pattern flanking the transcription start site, with a nucleosome-depleted dip at the centre. Profiles above each heatmap are the mean across all rows.
Figure 7 Replicate concordance (IDR) simulated
Figure 7 | Reproducible peaks. (a) Peak signal in replicate 1 versus replicate 2; green points pass an IDR-style concordance filter, grey points do not. (b) Peak counts used to compute the rescue and self-consistency ratios; both fall well below the ENCODE4 limit of 2.
Figure 8 Where peaks land case study 1
~34% of high-confidence peaks fall in promoter regions in Case Study 1.
The remainder are distributed across exons, introns and intergenic space, typical of distal regulatory elements.
Figure 8 | Genomic feature distribution. Annotation relative to nearest genes (promoter, exon, intron, intergenic) shows whether a factor behaves like a promoter-proximal or enhancer-oriented regulator, and is compared against the genomic background.
Comparison & Integration

Compare conditions, integrate marks

Differential binding shows what changes between conditions. Combining several histone marks reveals the chromatin state of each region: active promoter, enhancer, transcribed gene or Polycomb-silenced domain.

Figure 9 Differential binding between two conditions simulated
Figure 9 | Differential occupancy. (a) Volcano plot of log2 fold change against significance across a consensus peak set. (b) MA plot of fold change against mean signal; noise is larger at low signal, which count-based models account for. Sites with FDR < 0.05 and |log2FC| > 1 are highlighted: gained in green, lost in orange. When a treatment changes a mark globally, spike-in chromatin from another species provides the scaling reference that conventional normalisation would otherwise erase.
Figure 10 Multi-mark chromatin-state map of an example locus simulated
Figure 10 | From marks to states. H3K4me3 and H3K27ac mark the active promoter of Gene A; H3K4me1 with H3K27ac flags a distal enhancer; H3K36me3 spreads across the transcribed gene body; H3K27me3 covers the silenced Gene B in a broad Polycomb domain. Segmentation tools such as ChromHMM learn these recurring combinations to label every region of the genome.
Deliverables

Structured outputs, ready for biology

At project completion, clients receive a structured set of sequencing, analysis and reporting outputs via secure shared storage.

Raw & Processed Data

FASTQ files — raw and appropriately processed sequence data.

QC Reports

Sample-level QC including library complexity and enrichment metrics.

Alignment Files

Processed BAM files with associated index files.

Peak Calls

Narrow- or broad-peak files (BED/narrowPeak) per sample and condition.

Annotated Peaks

Peak sets annotated with nearest genes and genomic features.

Motif Results

Known and de novo motif enrichment outputs within called peaks.

Summary Report

Sample-wise QC, alignment, peak-calling and annotation statistics.

Secure Delivery

Project files via secure shared storage or cloud-based access.

Sample Requirements

What we accept

Flexible input options for sequencing-plus-analysis or analysis-only projects.

ChIP DNA & Controls

We accept chromatin immunoprecipitated DNA (ChIP DNA) and matched input/control DNA, as well as pre-generated sequencing data where analysis-only services are required.

Project-specific requirements for DNA quantity, concentration, library preparation, sequencing depth and read configuration are finalised according to the target protein, study design and sequencing strategy.

Contact us for a project-specific sample submission checklist and recommended sequencing depth.

ChIP DNA

ChIP DNA + matched input/control
or pre-generated FASTQ

Case Studies & Performance

Proven on real chromatin datasets

Production metrics from transcription-factor and histone-modification ChIP-Seq projects.

Transcription Factor Binding-Site Mapping

Narrow-peak calling · Matched input control

Mean raw-data Q30 92.1%
Mean primary alignment rate 98.6%
Mean FRiP score 0.09
Peaks called (high-confidence) ~18,400
Peaks annotated to promoters ~34%

The dataset achieved high sequencing quality and alignment rates, with high-confidence peak sets and annotated outputs delivered alongside a project summary report.

Histone Modification Profiling

Broad-peak calling · Replicate cohorts

Peak mode Broad domain
Enrichment metrics Consistent
Replicate reproducibility High
Differential binding Supported
Domain calls Reproducible

In production histone ChIP-Seq datasets, we have observed consistent enrichment metrics and reproducible broad-domain peak calls across biological replicates, supporting robust differential binding analysis where required.

The Frontier

Where chromatin profiling is heading

ChIP-seq remains the reference method, but a new generation of in-situ assays and sequence-based AI models is changing how regulatory genomes are mapped and interpreted. Here is how they relate.

Figure 11 How well do newer assays recover ChIP-seq signal? published data

Sensitivity for H3K4me3 peaks in mouse haploid round spermatids

ChIP-seq
0.99
CUT&Tag
0.99
CUT&RUN
0.74
≈54% of ENCODE ChIP-seq peaks for H3K27ac and H3K27me3 in K562 cells were recovered on average by combined CUT&Tag datasets, and the peaks recovered were the strongest ones.
Figure 11 | Benchmarks from 2025 studies. Bars show reported sensitivity for H3K4me3 (Wang et al., 2025). The same study found CUT&Tag gave the highest signal-to-noise for H3K4me3 and CTCF, but warned that its bias toward accessible chromatin may create promoter false positives. The callout summarises Abbasova et al., 2025. Different targets and cell types can behave differently, so choose the method by target.
Table 2 Choosing a chromatin-profiling assay
AssayPrincipleStrengthsWatch-outs
ChIP-seqCrosslinked chromatin immunoprecipitationReference standard with community QC rules and public dataNeeds more cells; higher background; antibody-dependent
CUT&RUNAntibody-guided nuclease cleaves DNA in situLow background; small inputsCan be weaker for some TFs; accessibility bias
CUT&TagAntibody-guided Tn5 tagmentation in situHigh signal-to-noise; very low input; single-cell formatsBias to open chromatin; recovers a subset of ChIP-seq peaks
ChIP-exo / nexusExonuclease trimming to the crosslink boundaryNear-base-pair footprintsMore complex protocol and library requirements
Table 2 | Assay comparison. Summarised from the literature cited in the references; the right choice depends on target class, sample availability and the claim the data must support.
Figure 12 Deep learning on epigenomic data: base-resolution sequence-to-function models schematic
ACGT DNA sequence ~1–2 kb, one-hot encoded Convolutionsmotif detectors Dilated residuallong-range motif syntax Profile head base-resolution read-count shape Counts head total signal in the window Bias model (ChromBPNet) What you learn • Contribution scores per base• Motifs and their spacing rules• Footprints of bound factors• Effect of any variant, in silico(prioritise non-coding variants)
Figure 12 | Sequence-to-function networks. BPNet (Avsec et al., Nat Genet 2021) was trained on ChIP-nexus data to predict base-resolution binding profiles and reveal flexible motif syntax. ChromBPNet extends the idea to chromatin accessibility by first learning, then freezing, a bias model of enzyme sequence preference so that the second network captures regulatory signal only; it is competitive with the much larger Enformer on variant benchmarks. Newer long-context models such as AlphaGenome predict many regulatory tracks from long DNA sequences to score variant effects.
IN SITU ASSAYS

CUT&RUN and CUT&Tag

Antibody-guided nuclease or transposase acts inside intact nuclei, which lowers background and allows profiling from small samples, including single cells. Peaks should be interpreted with the assay's accessibility bias in mind.

SINGLE CELL

Multimodal single-cell chromatin

Single-cell CUT&Tag and nanobody-based variants profile histone marks cell by cell, and some formats read several chromatin features in the same cells, resolving regulatory heterogeneity within tissues.

QUANTITATION

Spike-in normalisation

A fixed amount of foreign chromatin and a species-specific antibody target let reads be scaled to an internal reference, so global increases or decreases in a histone mark become measurable.

Scope note. Nucleome's ChIP-Seq service covers the workflow and deliverables described above. The frontier methods on this page are shown to place the service in its scientific context; contact our team to discuss extended or project-specific analyses.
Why Nucleome

Service highlights

End-to-end support for research and translational chromatin-biology projects.

End-to-End Analysis

From raw sequencing data through QC, alignment, peak calling, annotation and reporting.

Narrow & Broad Peaks

Analysis suited to both transcription-factor and histone-modification binding profiles.

Comprehensive Annotation

Peak annotation relative to genes and genomic features, with motif enrichment analysis.

Transparent QC

Library complexity, enrichment and FRiP metrics incorporated into project reporting.

Flexible Study Design

Support for individual samples, replicate cohorts and differential binding projects.

Secure Data Delivery

Raw, intermediate and final datasets delivered through secure project storage.

Questions, Basic To Advanced

Frequently asked, in plain language

basicWhat is ChIP-Seq in one sentence?
It is a method that finds every place in the genome where a chosen protein, or a chosen chemical mark on histones, is attached, by pulling that protein out of chromatin with an antibody and sequencing the DNA it was holding.
basicDo I need to do the sequencing as well?
No. We accept ChIP DNA with matched input for sequencing and analysis, and we also analyse sequencing data you already have, so you can start from either point.
basicWhy is an input control needed?
Open or repetitive regions can look enriched for reasons that have nothing to do with your target. The input sample shows that background, so peak callers report only regions enriched over it.
intermediateHow much sequencing depth should I plan?
As a community reference, ENCODE4 asks for at least 10 million usable fragments per replicate for transcription factors and narrow histone marks (more than 20 million recommended), and at least 20 million for broad marks (more than 45 million recommended). Your final depth depends on the target, genome size and study design, so our team confirms it with you for each project.
intermediateMy FRiP is low. Has the experiment failed?
Not necessarily. FRiP depends on the target, antibody and peak caller, and ENCODE4 reports it for comparison between similar experiments rather than as a pass/fail cut-off. Read it together with cross-correlation (NSC, RSC), library complexity and replicate concordance before judging the data.
intermediateNarrow-peak or broad-peak calling: which should I use?
Use narrow mode for sequence-specific transcription factors and sharp marks such as H3K4me3 or H3K27ac, and broad mode for domain-like marks such as H3K27me3 or H3K36me3. Mixed-mark studies can use both within one project.
advancedShould I use CUT&Tag or CUT&RUN instead of ChIP-seq?
They are strong choices for small samples and low background, but benchmarks show they are not simple drop-in replacements. A 2025 study found combined CUT&Tag data recovered about half of ENCODE H3K27ac and H3K27me3 peaks in K562 cells, while another found method performance varied by target. Pick the assay by target and by the claim the data must support, and validate surprising peaks.
advancedHow are differential binding results kept reliable?
Replicate-aware count models work on a consensus peak set, with normalisation chosen to match the biology. If a treatment is expected to change a mark globally, a spike-in reference is recommended, because normalising to total reads would hide the change. Results are reported with FDR control.
advancedCan ChIP-seq data feed machine-learning models?
Yes. Base-resolution models such as BPNet learn motif syntax from binding profiles, and related models predict how sequence variants alter regulatory activity. They depend on well-controlled, replicated data, which is why the QC standards on this page matter even more for modelling projects.
References

Key literature & standards

Selected sources that underpin the reference values and methods on this page. Figures labelled "simulated" are illustrations generated for this page, not client data.

  1. Landt SG, et al. ChIP-seq guidelines and practices of the ENCODE and modENCODE consortia. Genome Research 2012;22:1813–1831. Link
  2. ENCODE Project. ENCODE4 transcription factor ChIP-seq data standards and processing pipeline. encodeproject.org
  3. ENCODE Project. ENCODE4 histone ChIP-seq data standards and processing pipeline. encodeproject.org
  4. Zhang Y, et al. Model-based analysis of ChIP-Seq (MACS). Genome Biology 2008;9:R137.
  5. Li Q, Brown JB, Huang H, Bickel PJ. Measuring reproducibility of high-throughput experiments. Annals of Applied Statistics 2011;5:1752–1779.
  6. Amemiya HM, Kundaje A, Boyle AP. The ENCODE Blacklist: identification of problematic regions of the genome. Scientific Reports 2019;9:9354.
  7. Ross-Innes CS, et al. Differential oestrogen receptor binding is associated with clinical outcome in breast cancer. Nature 2012;481:389–393 (DiffBind).
  8. Orlando DA, et al. Quantitative ChIP-Seq normalization reveals global modulation of the epigenome. Cell Reports 2014;9:1163–1170.
  9. Heinz S, et al. Simple combinations of lineage-determining transcription factors prime cis-regulatory elements required for macrophage and B cell identities. Molecular Cell 2010;38:576–589 (HOMER).
  10. Ernst J, Kellis M. ChromHMM: automating chromatin-state discovery and characterization. Nature Methods 2012;9:215–216.
  11. Rhee HS, Pugh BF. Comprehensive genome-wide protein-DNA interactions detected at single-nucleotide resolution. Cell 2011;147:1408–1419 (ChIP-exo).
  12. He Q, Johnston J, Zeitlinger J. ChIP-nexus enables improved detection of in vivo transcription factor binding footprints. Nature Biotechnology 2015;33:395–401.
  13. Skene PJ, Henikoff S. An efficient targeted nuclease strategy for high-resolution mapping of DNA binding sites. eLife 2017;6:e21856 (CUT&RUN).
  14. Kaya-Okur HS, et al. CUT&Tag for efficient epigenomic profiling of small samples and single cells. Nature Communications 2019;10:1930.
  15. Abbasova L, et al. CUT&Tag recovers up to half of ENCODE ChIP-seq histone acetylation peaks. Nature Communications 2025;16. Link
  16. Wang R, et al. Benchmark of chromatin–protein interaction methods in haploid round spermatids. Frontiers in Cell and Developmental Biology 2025. Link
  17. Avsec Ž, et al. Base-resolution models of transcription-factor binding reveal soft motif syntax. Nature Genetics 2021;53:354–366. PubMed
  18. Avsec Ž, et al. Effective gene expression prediction from sequence by integrating long-range interactions (Enformer). Nature Methods 2021;18:1196–1203.
  19. ChromBPNet: bias factorized, base-resolution deep learning models of chromatin accessibility reveal cis-regulatory sequence syntax, transcription factor footprints and regulatory variants. bioRxiv 2024. Link
  20. Advancing regulatory variant effect prediction with AlphaGenome. 2026. PubMed