CLIP-seq Explained: How Crosslinking and Immunoprecipitation Map RNA-Binding Protein Sites

Short answer

CLIP-seq (crosslinking and immunoprecipitation followed by sequencing) maps where a specific RNA-binding protein sits on RNA. Cells are irradiated with UV light so the protein becomes covalently attached to its RNA, the protein is pulled down with an antibody, and the RNA fragments it held are sequenced. The reads form peaks that mark binding sites.

Definition and scope

CLIP-seq is the RNA counterpart of ChIP-seq. Where ChIP-seq asks which DNA a protein occupies, CLIP-seq asks which RNA a protein contacts, and where on that RNA. The name covers a family of related protocols rather than one method; the best-documented members are PAR-CLIP, iCLIP and eCLIP, compared below. The shared logic is the same in each: covalent UV crosslinking freezes a protein-RNA contact in the cell, so the immunoprecipitation can use stringent washes without losing the interaction.

Because the antibody does the selecting, a CLIP-seq experiment reports on one protein at a time. The co-IP controls guide covers the antibody and lysis considerations that carry over.

How CLIP-seq works in practice

The published eCLIP protocol (Van Nostrand et al., 2016, Nature Methods) shows the common sequence of events:

• Crosslink. Cells are exposed to UV light at 254 nm (400 mJ/cm² in the eCLIP paper), which attaches proteins to the RNA they touch.
• Trim. The lysate is partially digested with RNase I so each protein keeps only a short RNA fragment.
• Immunoprecipitate. An antibody on magnetic beads captures the target protein and its RNA; stringent washes remove non-specific material.
• Add adapters. In eCLIP a barcoded RNA adapter is ligated on the beads and a second DNA adapter after reverse transcription.
• Sequence. Reverse transcription usually stops at the crosslink site, so the start of each read points to the contact position, giving single-nucleotide resolution. An inline random sequence in the adapter lets analysts tell true PCR duplicates from independent fragments.

Reads then enter the usual sequencing pipeline: FASTQ files are aligned to give BAM files, and peaks are reported as intervals in BED coordinates. The eCLIP paper reports that the protocol omits radiolabeling and can need as little as 4 days of hands-on time.

PAR-CLIP, iCLIP and eCLIP compared

The variants differ in how the crosslink is made and how the crosslink position is read.

Variant	Crosslinking	How the site is read	Reported performance
PAR-CLIP	Cells grown with 4-thiouridine (4SU), irradiated at 365 nm	T-to-C transitions in reads, because reverse transcription incorporates dG instead of dA at the crosslink	Photoactivatable nucleosides improved RNA recovery 100- to 1000-fold at the same radiation energy (Hafner et al., 2010)
iCLIP	UV light	Reverse-transcription truncation, using circular ligation	28 PCR cycles in the paper’s comparison; median 13.9% usable reads (Van Nostrand et al., 2016)
eCLIP	UV 254 nm	Reverse-transcription truncation, using a second adapter ligated after reverse transcription	16 PCR cycles; median 84.1% usable reads (Van Nostrand et al., 2016)

Two points stand out. First, PAR-CLIP needs a nucleoside that cells can take up: the 4SU-labeled cells are the price of the diagnostic mutation. Without a diagnostic mutation, as in 254 nm datasets, the PAR-CLIP authors state the exact crosslink position could not be readily deduced from mutations, which are at least one order of magnitude less frequent than the T-to-C transitions seen in PAR-CLIP. Second, the eCLIP authors state that circular ligation as used in iCLIP is often inefficient, and report that eCLIP decreases the required amplification by about 1,000-fold, with a usable fraction of 85.2% against 30.8% for iCLIP at 20 million mapped reads.

What makes a CLIP-seq dataset credible

The ENCODE Consortium publishes data standards for eCLIP. They double as a checklist for reading any CLIP-seq paper or dataset:

• Replicates. Two or more biological replicates.
• Input control. A size-matched input control for each experiment, with matching run type and read length.
• Antibody. An antibody characterized to ENCODE Consortium standards.
• Depth. 1 million unique fragments, or saturated peak detection, in each biological replicate.
• Enrichment. For narrow-binding RNA-binding proteins, a fraction of reads in peaks (FRiP) of at least 0.005 over enriched peaks, with exemptions possible for atypical binders.
• Replicate agreement. The experiment passes IDR concordance if both the rescue and self-consistency ratios are below 2.

Why the size-matched input matters

In eCLIP the size-matched input (SMInput) is 2% of the sample taken after lysis and fragmentation but before immunoprecipitation, processed identically, including the same membrane size selection. Peaks are normalised against it to cut false positives. A peak list produced without such a control cannot separate protein-specific signal from background that is abundant in the starting material.

Common misunderstandings

• “CLIP-seq is one protocol.” It is a family. PAR-CLIP depends on 4SU labeling and 365 nm light; eCLIP uses 254 nm light and reads the reverse-transcription stop point.
• “An IP-only peak list is enough.” The ENCODE standards require a size-matched input for every eCLIP experiment.
• “Higher read counts mean a better library.” The eCLIP paper compares usable fractions, not just depth: at 20 million mapped reads 85.2% of eCLIP reads were usable against 30.8% for iCLIP.
• “A peak proves function.” The standards above test reproducibility, controls and depth. They do not test whether binding changes the fate of the RNA.

FAQ

What does CLIP-seq stand for?

Crosslinking and immunoprecipitation followed by sequencing. It identifies where a specific RNA-binding protein contacts RNA.

What is the difference between PAR-CLIP, iCLIP and eCLIP?

PAR-CLIP uses 4-thiouridine labeling and 365 nm light and reads T-to-C transitions. iCLIP and eCLIP use UV crosslinking and read the reverse-transcription stop; eCLIP ligates a second adapter after reverse transcription instead of using circular ligation.

Why is a size-matched input control needed?

It is processed like the immunoprecipitated sample but without the antibody pull-down, so peaks can be normalised against background. ENCODE requires one for every eCLIP experiment.

How many replicates does ENCODE expect?

Two or more biological replicates, each with 1 million unique fragments or saturated peak detection.

Does a CLIP-seq peak prove the protein regulates that RNA?

No. The data standards test reproducibility, controls and depth, not whether binding changes the fate of the RNA.

Sources

Primary sources checked on 2026-10-29.

• Van Nostrand EL et al. (2016). Robust transcriptome-wide discovery of RNA-binding protein binding sites with enhanced CLIP (eCLIP). Nature Methods. https://pmc.ncbi.nlm.nih.gov/articles/PMC4887338/. Checked 2026-10-29.
• Hafner M et al. (2010). Transcriptome-wide identification of RNA-binding protein and microRNA target sites by PAR-CLIP. Cell. https://pmc.ncbi.nlm.nih.gov/articles/PMC2861495/. Checked 2026-10-29.
• ENCODE Consortium. eCLIP data standards. https://www.encodeproject.org/eclip/. Checked 2026-10-29.