Do I need an IgG control for CUT&RUN or CUT&Tag?

You can run one as an antibody specificity check, but it is not what makes peak calling honest the way an input control is for ChIP-seq. Because background is near zero, spike-in DNA carried over by the tethered enzyme is the standard way to normalize across samples instead.

Can I just run MACS2 on CUT&RUN or CUT&Tag data the same way I would on ChIP-seq?

You can run MACS2, but its defaults were tuned for ChIP-seq's higher background, so they will not behave the same way with an IgG control on low-background data. Many labs switch to SEACR instead, which is built for this style of data and takes bedGraph input from paired-end reads.

Is CUT&Tag always a better choice than CUT&RUN?

No. CUT&Tag tends to give sharper peaks and higher FRiP for histone marks and many protein targets, but it can underperform CUT&RUN for certain transcription factors and chromatin-binding proteins, possibly because Tn5 does not access every epitope as reliably as protein A/G-MNase. Check the literature for your specific target before assuming CUT&Tag is the upgrade.

How many cells do I actually need for each method?

ChIP-seq typically needs millions of cells. CUT&RUN needs roughly 50,000 to 300,000 cells, and CUT&Tag can work with as few as 5,000 to 50,000, which is the main reason both have replaced ChIP-seq for rare or FACS-sorted populations.

Can I compare new CUT&RUN peaks against an old ChIP-seq dataset from GEO?

Only after accounting for differences in background, control strategy, and peak-caller settings between the two assays; a raw peak count or overlap comparison without matching genome build, normalization, and thresholds will mislead you. Reprocessing the GEO dataset through a consistent pipeline is safer than comparing peaks called years apart with different defaults.

Comparison · assay

CUT&RUN and CUT&Tag vs ChIP-seq: Which One Should You Use?

Cell number, background level, and peak-caller defaults diverge enough that swapping one assay for the other without retuning your pipeline hands you peaks that are not real.

By Ming "Tommy" Tang, Director of Bioinformatics in Big Pharma · Reviewed September 2026 · 4 min read

The verdict

Default to CUT&RUN or CUT&Tag for new experiments, especially when cell number is limited or you want cleaner peaks with less sequencing. CUT&Tag is the better default of the two for histone marks and most protein targets because it gives the sharpest peaks and the highest FRiP; reach for CUT&RUN instead when your target is a transcription factor or chromatin-binding protein that CUT&Tag is known to underperform on, since Tn5 tagmentation does not access every epitope the way protein A/G-MNase does.

Stick with ChIP-seq when you are reprocessing existing public data and need peaks comparable to that corpus, when your antibody has years of ChIP-seq validation but unknown performance in native, non-crosslinked conditions, or when your lab's whole downstream pipeline, MACS2 with cross-correlation QC, input controls, and established broad and narrow settings, is already built around ChIP-seq assumptions. Do not port a ChIP-seq MACS2 command onto CUT&RUN or CUT&Tag data unchanged; the near-zero background breaks the assumptions those defaults were tuned for.

ChIP-seq crosslinks protein to DNA, fragments bulk chromatin, then uses an antibody to pull the target complex out of the lysate. Background comes from everywhere the antibody sticks nonspecifically across the whole genome, which is why the field built peak callers around a matched input or IgG control from the start.

CUT&RUN and CUT&Tag skip the pulldown. An antibody bound to its target inside an intact or lightly permeabilized cell recruits a tethered enzyme, protein A/G-MNase for CUT&RUN or protein A/G-Tn5 for CUT&Tag, that cuts or tags DNA only in the immediate vicinity of the epitope. Almost nothing gets cut anywhere else, so the background that ChIP-seq spends an input control correcting for barely exists here.

That single mechanistic difference cascades into everything downstream: how many cells you need, whether you trust an input control or a spike-in, whether fragment size is noise or signal, and which peak caller settings you can safely default to.

Head to head

Criterion	CUT&RUN and CUT&Tag	ChIP-seq	Edge
Minimum cell input	CUT&RUN needs roughly 50,000 to 300,000 cells; CUT&Tag needs as few as 5,000 to 50,000 cells.	ChIP-seq typically needs millions of cells to recover enough DNA-protein complex for antibody pulldown.	CUT&RUN and CUT&Tag
Signal-to-noise (FRiP)	CUT&Tag produces the highest FRiP values of the three methods; CUT&RUN's FRiP is lower with broader peak distributions.	ChIP-seq FRiP depends heavily on target enrichment level, so antibody quality swings it more than the protocol itself.	CUT&RUN and CUT&Tag
Peak shape and resolution	CUT&Tag gives sharp, focal peaks at both histone marks and protein binding sites; CUT&RUN gives broader domains with wider spread around the target.	ChIP-seq peak shape is target-dependent, sharp for transcription factors and broad for marks like H3K27me3 and H3K36me3, and MACS2 needs different settings for each.	Tie
Control and normalization strategy	Background is near zero, so spike-in DNA carried over by the tethered enzyme is the standard way to normalize across samples instead of an input control.	Peak calling assumes a matched input or IgG control; without one, copy-number variation and open-chromatin bias masquerade as real peaks.	Tie These are different normalization problems, not a strictly better or worse axis.
Transcription factor performance	CUT&RUN can outperform CUT&Tag on some chromatin-binding proteins; CUT&Tag underperforms on certain TFs, possibly due to limited Tn5 access to the epitope.	ChIP-seq has decades of antibody validation data for TFs, including antibodies that were never tested in native, non-crosslinked conditions.	Tie
Peak caller maturity	CUT&RUN and CUT&Tag increasingly rely on specialized callers like SEACR, which takes bedGraph files built from paired-end reads.	MACS2 is the long-standing default for ChIP-seq, with documented narrow (q < 0.01) and broad (--broad, q < 0.05) modes and an extsize tied to cross-correlation fragment length.	ChIP-seq
QC metric standardization	QC for CUT&RUN and CUT&Tag still borrows ChIP-seq metrics like FRiP and duplicate rate, without an equally standardized peak-free check.	ChIP-seq has an established peak-call-free QC step: strand cross-correlation, reported as PPQT-NSC, robust to depth, and PPQT-RSC, sensitive to depth.	ChIP-seq
Fragment size as a biological signal	Fragment-size classes carry information about what was cut, so paired-end sequencing and fragment-level analysis are required, not optional.	Fragment size mainly feeds the extsize parameter for peak calling and is rarely treated as a standalone biological readout.	CUT&RUN and CUT&Tag
Sequencing depth per library	Low background means libraries need only a few million reads to call peaks confidently.	Higher genome-wide background means deeper sequencing is needed to separate real enrichment from noise.	CUT&RUN and CUT&Tag
Duplicate rate interpretation	Data is typically filtered to MAPQ 30 or higher, concordant pairs only, then peak-called with duplicates kept rather than blanket deduplication.	Duplicate rate is target-dependent, from 51 to 62 percent for ER peaks up to over 90 percent for NRF1 and H3K4me3, reflecting enrichment strength rather than PCR artifact alone.	Tie

Use CUT&RUN and CUT&Tag when

You have limited starting material, such as a FACS-sorted rare population, a small biopsy, or anything under a few hundred thousand cells.
You are profiling a histone mark or protein target where sharp peak resolution and high FRiP matter more than matching an existing ChIP-seq atlas.
You want lower sequencing cost per sample, since low background means these libraries need only a few million reads to call peaks confidently.
You are running a native, non-crosslinked chromatin profiling experiment and want to avoid crosslinking artifacts altogether.
A prior ChIP-seq attempt on this target failed or gave weak enrichment in bulk chromatin, and you suspect the problem was signal, not biology.

Use ChIP-seq when

You need peaks that are directly comparable to a large existing ChIP-seq corpus like ENCODE for cross-study analysis.
Your antibody has years of ChIP-seq validation but no published performance data in native, non-crosslinked CUT&RUN or CUT&Tag conditions.
You are reprocessing or reanalyzing existing public ChIP-seq datasets from GEO and need to stay within an established pipeline.
You are profiling an interaction that depends on crosslinking to capture, such as a transient or weak protein-DNA contact.
Your lab or core facility already has millions of cells available and a MACS2-based pipeline with cross-correlation QC built and validated.

Switching between them

Moving from a ChIP-seq pipeline to CUT&RUN or CUT&Tag changes more than the wet lab protocol. Input files change shape: SEACR, a common CUT&Tag caller, wants bedGraph files built from paired-end reads, not the BAM-based workflow MACS2 expects. Filtering changes too: CUT&RUN and CUT&Tag BAMs are typically restricted to MAPQ 30 or higher and concordant pairs only before peak calling. Normalization changes fundamentally: swap input subtraction for spike-in normalization using the E. coli DNA the enzyme carries over, since there is no meaningful input signal left to subtract. Thresholds do not carry over either, a MACS2 q-value that calls a sane number of peaks on ChIP-seq data can call far too many or far too few on CUT&RUN or CUT&Tag data with its much lower background, so treat q-value as something to tune against known positive and negative regions in IGV, not a fixed setting you copy between assays.

Pitfalls with either

Running MACS2 with default settings and an IgG control on CUT&RUN or CUT&Tag data and expecting ChIP-seq peak counts is a mistake; adjust the q-value or switch to SEACR, and validate against known positive regions instead of trusting defaults.
Treating spike-in normalization as optional for CUT&RUN or CUT&Tag is a mistake, since background is near zero and cell number varies between samples; spike-in DNA from the tethered enzyme is the correct way to normalize, not library size.
Applying a blanket duplicate-removal step to ChIP-seq BAMs without checking the target is a mistake, because duplicate rate is target-dependent, from 51 to 62 percent for ER peaks up to over 90 percent for NRF1 and H3K4me3, so high duplication near real peaks is not automatically a PCR artifact.
Sequencing CUT&RUN or CUT&Tag libraries single-end throws away information, since fragment-size classes carry biological signal about what was cut; paired-end sequencing and fragment-level analysis are required, not optional.
Using the narrow-peak default of q less than 0.01 for broad marks like H3K27me3 or H3K36me3 under-calls real domains; use --broad with q less than 0.05 for two-level calling, and expect CUT&RUN to give broader domains than CUT&Tag even for the same mark.
Assuming CUT&Tag is a strict upgrade over CUT&RUN for every target is a mistake, since CUT&Tag can underperform for certain transcription factors and chromatin-binding proteins; check prior literature or run a small pilot before switching an established TF protocol.

Questions people ask

Do I need an IgG control for CUT&RUN or CUT&Tag?: You can run one as an antibody specificity check, but it is not what makes peak calling honest the way an input control is for ChIP-seq. Because background is near zero, spike-in DNA carried over by the tethered enzyme is the standard way to normalize across samples instead.
Can I just run MACS2 on CUT&RUN or CUT&Tag data the same way I would on ChIP-seq?: You can run MACS2, but its defaults were tuned for ChIP-seq's higher background, so they will not behave the same way with an IgG control on low-background data. Many labs switch to SEACR instead, which is built for this style of data and takes bedGraph input from paired-end reads.
Is CUT&Tag always a better choice than CUT&RUN?: No. CUT&Tag tends to give sharper peaks and higher FRiP for histone marks and many protein targets, but it can underperform CUT&RUN for certain transcription factors and chromatin-binding proteins, possibly because Tn5 does not access every epitope as reliably as protein A/G-MNase. Check the literature for your specific target before assuming CUT&Tag is the upgrade.
How many cells do I actually need for each method?: ChIP-seq typically needs millions of cells. CUT&RUN needs roughly 50,000 to 300,000 cells, and CUT&Tag can work with as few as 5,000 to 50,000, which is the main reason both have replaced ChIP-seq for rare or FACS-sorted populations.
Can I compare new CUT&RUN peaks against an old ChIP-seq dataset from GEO?: Only after accounting for differences in background, control strategy, and peak-caller settings between the two assays; a raw peak count or overlap comparison without matching genome build, normalization, and thresholds will mislead you. Reprocessing the GEO dataset through a consistent pipeline is safer than comparing peaks called years apart with different defaults.

Glossary · Q-value
Glossary · Snakemake

Sources

Comparative analyses of ChIP-seq, CUT&RUN and CUT&Tag for Polycomb chromatin profiling — Source for FRiP, peak shape, and transcription factor performance comparisons across the three assays.
MACS2 peak calling details and parameters — Source for narrow vs broad q-value defaults and the extsize parameter.
Theoretical characterisation of strand cross-correlation in ChIP-seq — Source for the strand cross-correlation QC method used in ChIP-seq.
Identification of factors associated with duplicate rate in ChIP-seq data — Source for target-dependent duplicate rate figures.
CUT&RUNTools 2.0: A pipeline for single-cell and bulk-level CUT&RUN and CUT&Tag data analysis — Source for the SEACR peak caller description.