Analyzing m6A-seq / MeRIP Data

Before you start

You've done Epitranscriptomics 101 (what m6A is) and the NGS workflow (alignment, peak calling).
This is the analysis backbone of the Track 4 deliverable. It pairs naturally with RNA-seq from Track 3.
Any term new? The Glossary has it.

Learning objectives

By the end of this lesson you will be able to: describe how MeRIP-seq pulls down methylated RNA, walk through the peak-calling analysis workflow, explain the resolution problem and how the field improved on it, and read an m6A result critically.

MeRIP-seq: pull down the methylated RNA

MeRIP-seq (methylated RNA immunoprecipitation sequencing), also called m6A-seq, was the breakthrough that first mapped m6A across the transcriptome in 2012. The idea is elegant: fragment the RNA into roughly 100-nucleotide pieces, then use an anti-m6A antibody to pull down only the fragments that carry the modification. Sequence those, and the methylated regions pile up.

But a pulldown alone is not enough, because a highly expressed gene contributes more fragments to everything, methylated or not. So every MeRIP experiment is run as a pair: the antibody pulldown (IP) and a matched input sample (the same fragmented RNA, not immunoprecipitated, which is essentially a standard RNA-seq library). m6A sites are regions enriched in the IP relative to the input.

Decode the jargonIP vs input

IP is the anti-m6A immunoprecipitation: fragments containing m6A. Input is the matched total RNA without pulldown, which captures how much of each transcript was there to begin with. Comparing IP to input lets you measure methylation, not just abundance.

The analysis workflow

QC and trim the IP and input FASTQ files (same as any RNA-seq).
Align both to the genome with a splice-aware aligner (STAR or HISAT2), because mRNA reads cross exon junctions.
Call peaks: find regions significantly enriched in IP over input. Purpose-built tools model this exome-aware: exomePeak2, MeTPeak, or MACS2 adapted for RNA.
QC the peaks: a good m6A dataset recovers the DRACH motif under the peaks and shows enrichment near stop codons / 3' UTRs. If it does not, something is wrong.
Differential methylation between conditions (e.g. treatment vs control, or writer-knockdown vs control): tools like exomePeak2, RADAR, or QNB test whether the IP/input ratio changes, while accounting for expression changes.

# Conceptual shape (per condition you have IP + input, in replicates)
$ STAR --genomeDir hg38 --readFilesIn ip.fastq    --outSAMtype BAM ...
$ STAR --genomeDir hg38 --readFilesIn input.fastq --outSAMtype BAM ...
# In R/Bioconductor: peaks = IP enriched over input
> exomePeak2(bam_ip = "ip.bam", bam_input = "input.bam", gff = "annotation.gtf")

⚠️ The mistake that ruins m6A studies: confusing methylation with expression

If a gene's expression doubles between two conditions, its IP reads roughly double too, even with no change in methylation. A peak that looks "stronger" in raw IP counts may simply reflect more transcript. You must normalize the IP signal against the matched input (the IP/input ratio), and differential-methylation tools must control for expression changes. Reporting an "m6A increase" that is really an expression increase is a classic, avoidable error.

The resolution problem, and how the field fixed it

Antibody MeRIP has a real limitation: because you select ~100 nt fragments, a peak tells you m6A is somewhere in this ~100 nt window, not which adenosine. For mechanistic work you often need the exact base. Several methods deliver single-nucleotide or antibody-free maps:

Method	Approach	What it adds
miCLIP / m6A-CLIP	UV-crosslink the anti-m6A antibody to RNA, then read the crosslink signature (like CLIP).	Single-nucleotide resolution from an antibody method.
GLORI	Antibody-free chemical conversion that deaminates unmethylated A but spares m6A, read out as a base change.	Quantitative, single-base, high sensitivity; no antibody bias.
Nanopore direct RNA	Sequence native RNA directly; m6A perturbs the ionic current signal and can be detected.	No antibody, no chemical conversion; sees modifications on full-length native molecules.

Each trades off cost, sensitivity, and complexity. MeRIP remains the most common entry point, but a credible modern study often pairs it with a single-nucleotide method to nail down exact sites.

🔶 Level up: practice on real data (the Track 4 deliverable)

Public m6A datasets live in GEO (search "MeRIP-seq" or "m6A-seq"). A strong capstone for this track: take one MeRIP dataset, call peaks with exomePeak2, confirm the DRACH motif and stop-codon enrichment as QC, and write up which transcripts are methylated. That is a portfolio-grade epitranscriptomics analysis.

Check your understanding

Between control and drug-treated cells, gene X shows twice as many m6A IP reads in the treated condition. A student concludes "the drug increases m6A on gene X." What must they check first?

Correct. IP read counts scale with transcript abundance. If gene X's expression doubled, its IP reads double with no change in methylation. The valid signal is a change in the IP/input ratio, which is exactly why MeRIP requires a matched input and why differential tools control for expression.

A reviewer asks you to report the exact methylated adenosine for a functional follow-up, but you only ran standard antibody MeRIP-seq. What is the honest, correct response?

Right. Fragment-based MeRIP localizes m6A to a ~100 nt window, not a specific nucleotide. To resolve the exact site you switch to a single-nucleotide approach (miCLIP/m6A-CLIP, GLORI, or nanopore direct RNA). Recognizing this limit is part of doing the analysis honestly.

Your MeRIP peaks do NOT recover the DRACH motif and are spread evenly across gene bodies rather than near stop codons. What is the most reasonable conclusion?

Exactly. Recovering the DRACH motif and the characteristic enrichment near stop codons / 3' UTRs are standard positive-control checks for m6A data. Their absence points to a technical issue (specificity, enrichment, or pipeline error) rather than a discovery, and should be investigated before any biological claim.

In a standard MeRIP-seq (m6A-seq) experiment, what does the IP sample contain compared with the input sample?

Correct. MeRIP fragments RNA, then immunoprecipitates fragments carrying m6A with an antibody (the IP). The input is the matched fragmented RNA without IP. Peaks are regions enriched in IP relative to input.

A standard antibody-based MeRIP-seq peak typically spans roughly a hundred or more nucleotides. What is the practical limitation this creates?

Right. MeRIP resolution is set by the RNA fragment size, so a peak marks a window rather than one base. To get single-nucleotide resolution you need methods such as miCLIP rather than standard MeRIP-seq.

Sources & further reading

Dominissini D, et al. Topology of the human and mouse m6A RNA methylomes revealed by m6A-seq. Nature, 2012. (The original MeRIP/m6A-seq method.)
Meng J, et al. exomePeak: a Bioconductor package for peak calling and differential analysis of MeRIP-seq and m6A-seq data. Methods, 2014.
Linder B, et al. Single-nucleotide-resolution mapping of m6A and m6Am throughout the transcriptome (miCLIP). Nat Methods, 2015.
Liu C, et al. Absolute quantification of single-base m6A methylation in the mammalian transcriptome using GLORI. Nat Biotechnol, 2023.

Last reviewed: June 2026.

Next in Track 4