# Bio Workflows Merip Pipeline

> Orchestrates an end-to-end MeRIP-seq / m6A-seq analysis from raw FASTQ to differential m6A peak calls and metagene plots, chaining fastp adapter trimming, STAR splice-aware alignment (NO deduplication for non-UMI MeRIP), deepTools replicate-concordance + IP-enrichment QC, PreSeq saturation curves, exomePeak2 (transcript-aware, GC-bias-aware negative-binomial GLM) peak calling, optional MACS3 broad-peak cross-check, DRACH motif confirmation as a sanity check (NOT a per-peak filter), exomePeak2 differential calling via the four-BAM-vector interface (bam_ip + bam_input control; bam_treated_ip + bam_treated_input treatment), ChIPseeker annotation, and the canonical Guitar metagene with stop-codon enrichment as the biological QC anchor. Use when running a complete MeRIP analysis from raw reads, when chaining the constituent epitranscriptomics skills (merip-preprocessing -> m6a-peak-calling -> m6a-differential -> modification-visualization), or when wrapping the pipeline in Snakemake / Nextflow.

- Skill: `pku-yuangroup/bio-workflows-merip-pipeline` (Agent Skill, multi-file: 3 files)
- Install (CLI): `npx skillmds@latest add pku-yuangroup/bio-workflows-merip-pipeline`
- Raw SKILL.md: https://api.skillmd.com/api/skills/pku-yuangroup/bio-workflows-merip-pipeline/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: DevOps & Infra
- Author: pku-yuangroup (https://skillmd.com/u/pku-yuangroup)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/pku-yuangroup/bio-workflows-merip-pipeline

---


## Version Compatibility

Reference examples tested with: STAR 2.7.11+, samtools 1.19+, fastp 0.23+, deepTools 3.5+, PreSeq 3.2+, exomePeak2 1.14.x (Bioconductor 3.18 ONLY -- deprecated in Bioc 3.19, removed in 3.20; on current Bioc install from the Bioc 3.18 archive or use a successor), MACS3 3.0+, ChIPseeker 1.38+, Guitar 2.18+, BSgenome.Hsapiens.UCSC.hg38 1.4+, TxDb.Hsapiens.UCSC.hg38.knownGene 3.18+, HOMER 4.11+.

Before using code patterns, verify installed versions match. If versions differ:
- R: `packageVersion('<pkg>')` then `?function_name` to verify parameters
- CLI: `<tool> --version` then `<tool> --help` to confirm flags

If code throws ImportError, AttributeError, or TypeError, introspect the installed package and adapt the example to match the actual API rather than retrying.

exomePeak2 has NO `mode=` or `experiment_design=` argument; differential is triggered by populating `bam_treated_ip` + `bam_treated_input`. MeTPeak defaults are `WINDOW_WIDTH=50, SLIDING_STEP=50, FRAGMENT_LENGTH=100`. MACS3 default `--keep-dup` is 1 and MUST be overridden to `all` for non-UMI MeRIP. Guitar `txTxdb=` is the modern argument name (older releases used `txdb=`).

# MeRIP-seq End-to-End Pipeline

**"Analyze my MeRIP-seq data from FASTQ to differential m6A peaks"** -> Orchestrate read alignment (STAR splice-aware to GENOME), IP-enrichment QC (deepTools plotFingerprint, replicate Spearman, PreSeq saturation), m6A peak calling (exomePeak2 transcript-aware default, MACS3 broad as cross-check), DRACH motif sanity check (HOMER), exomePeak2 differential via the four-BAM-vector interface, ChIPseeker feature annotation, and Guitar transcript-feature metagene confirming canonical stop-codon enrichment. Defer per-skill deep treatment to `epitranscriptomics/merip-preprocessing`, `epitranscriptomics/m6a-peak-calling`, `epitranscriptomics/m6a-differential`, and `epitranscriptomics/modification-visualization`.

This is a workflow skill: it owns the chaining decisions and hand-offs, not the internals of any one step.

## The governing principle

MeRIP-seq inverts several DNA-pipeline reflexes; the trustworthy callset is decided at these seams.

1. **Do NOT deduplicate non-UMI MeRIP — duplicates are signal, not artifact.** A highly methylated, highly expressed transcript legitimately produces many identical fragments; removing them (or MACS3 default `--keep-dup 1`) erases the strongest m6A peaks. Keep all reads (`--keep-dup all`); only dedup when a UMI is present.
2. **Enrichment is IP-vs-Input, and differential is a FOUR-BAM comparison.** Every condition needs its own IP AND Input. exomePeak2 has no `mode=`/`experiment_design=` argument; differential is triggered simply by populating `bam_treated_ip` + `bam_treated_input` alongside the control `bam_ip` + `bam_input`.
3. **DRACH is a peak-SET sanity check, never a per-peak filter.** Confirm the motif is enriched across the whole peak set (P-value < 1e-50); post-hoc dropping individual peaks that lack a DRACH match discards real non-canonical sites and biases the callset.
4. **The Guitar stop-codon metagene is the biological go/no-go.** m6A concentrates near the stop codon / 3'UTR-proximal CDS end; if that enrichment is absent, the IP failed or the antibody is wrong — STOP, do not interpret downstream. And before comparing peak COUNTS across conditions, rarefy BAMs to a common unique-read depth (peak number scales with depth).

## Pipeline Overview

```
FASTQ -> fastp trim -> STAR genome align -> samtools sort/index -> deepTools QC + PreSeq saturation
       -> exomePeak2 peak calling (+ MeTPeak / MACS3 cross-check)
       -> HOMER DRACH sanity check
       -> exomePeak2 differential (bam_ip + bam_treated_ip)
       -> ChIPseeker feature annotation
       -> Guitar metagene (stop-codon QC anchor) + pyGenomeTracks browser figures
```

## Step 1: Adapter Trimming

```bash
fastp \
    --in1 raw/IP_R1.fastq.gz --in2 raw/IP_R2.fastq.gz \
    --out1 trimmed/IP_R1.fq.gz --out2 trimmed/IP_R2.fq.gz \
    --json qc/IP_fastp.json --html qc/IP_fastp.html \
    --length_required 25 --detect_adapter_for_pe --thread 8

fastp \
    --in1 raw/Input_R1.fastq.gz --in2 raw/Input_R2.fastq.gz \
    --out1 trimmed/Input_R1.fq.gz --out2 trimmed/Input_R2.fq.gz \
    --json qc/Input_fastp.json --html qc/Input_fastp.html \
    --length_required 25 --detect_adapter_for_pe --thread 8
```

Standard non-UMI MeRIP: do NOT pass `--umi`. See `epitranscriptomics/merip-preprocessing` for the do-NOT-dedup rationale.

## Step 2: STAR Splice-Aware Genome Alignment

```bash
STAR --runMode alignReads \
    --genomeDir refs/star_index \
    --readFilesIn trimmed/IP_R1.fq.gz trimmed/IP_R2.fq.gz \
    --readFilesCommand zcat \
    --outSAMtype BAM SortedByCoordinate \
    --outFilterMultimapNmax 20 \
    --outSAMattributes NH HI AS nM NM MD \
    --outFileNamePrefix aligned/IP_rep1_ \
    --runThreadN 12

samtools index aligned/IP_rep1_Aligned.sortedByCoord.out.bam
ln -sf IP_rep1_Aligned.sortedByCoord.out.bam aligned/IP_rep1.bam     # downstream QC/peak steps consume the short ${sample}_rep${n}.bam name
ln -sf IP_rep1_Aligned.sortedByCoord.out.bam.bai aligned/IP_rep1.bam.bai
```

Repeat for each IP and Input replicate. Align to GENOME (not transcriptome) for downstream MeRIP peak calling. Do NOT deduplicate (no UMI in standard MeRIP).

## Step 3: IP-Enrichment + Replicate-Concordance QC

```bash
multiBamSummary bins \
    --bamfiles aligned/IP_rep[0-9].bam aligned/Input_rep[0-9].bam \
    --binSize 10000 --numberOfProcessors 8 \
    -o qc/cov.npz

plotCorrelation --corData qc/cov.npz --corMethod spearman --skipZeros \
    --whatToPlot heatmap --colorMap RdYlBu_r --plotNumbers \
    -o qc/replicate_correlation.pdf

plotFingerprint \
    --bamfiles aligned/IP_rep[0-9].bam aligned/Input_rep[0-9].bam \
    --skipZeros --numberOfProcessors 8 \
    --JSDsample aligned/Input_rep1.bam \
    --outQualityMetrics qc/fingerprint_metrics.tab \
    -o qc/fingerprint.pdf

preseq lc_extrap -B -o qc/IP_rep1_lc_extrap.txt aligned/IP_rep1.bam
```

For peak-count comparison across conditions, rarefy BAMs to a common unique-read depth informed by the saturation curve before calling peaks.

## Step 4: exomePeak2 Peak Calling (Per-Condition)

**Goal:** Produce a transcript-aware set of m6A peaks with FDR and IP/input fold-change from paired IP/Input genome BAM files, suitable as input to differential analysis, motif scanning, or downstream visualisation.

**Approach:** Build a TxDb from the matched GTF; pass paired IP/Input BAM vectors to `exomePeak2()` with `txdb` and `genome` (BSgenome) for GC correction; export BED12 + RDS to `save_dir/`.

```r
library(exomePeak2)
library(GenomicFeatures)
library(BSgenome.Hsapiens.UCSC.hg38)

txdb <- makeTxDbFromGFF('refs/annotation.gtf', format='gtf')

result <- exomePeak2(
    bam_ip       = c('aligned/IP_rep1.bam', 'aligned/IP_rep2.bam', 'aligned/IP_rep3.bam'),
    bam_input    = c('aligned/Input_rep1.bam', 'aligned/Input_rep2.bam', 'aligned/Input_rep3.bam'),
    txdb         = txdb,
    bsgenome     = BSgenome.Hsapiens.UCSC.hg38,   # bsgenome= (a BSgenome object) for GC correction; genome= would be the UCSC string 'hg38'
    paired_end   = TRUE,
    library_type = 'unstranded',
    save_dir     = 'exomePeak2_output'            # no experiment_name arg; output goes straight under save_dir/
)

peaks <- result
nrow(peaks)   # SummarizedExomePeak has no length method (would return 1); nrow = peak count
```

`exomePeak2()` writes fixed filenames under `save_dir/`: `Mod.bed` (BED12 peaks), `Mod.csv` (per-peak fold-change / FDR), `Mod.rds`.

## Step 5: MACS3 Broad-Peak Cross-Check (Optional)

```bash
macs3 callpeak \
    --treatment aligned/IP_rep[0-9].bam \
    --control aligned/Input_rep[0-9].bam \
    --format BAMPE --gsize hs \
    --nomodel --extsize 150 \
    --keep-dup all \
    --broad --broad-cutoff 0.1 --qvalue 0.05 \
    --outdir macs3_output --name m6a_run1
```

`--keep-dup all` is non-negotiable for non-UMI MeRIP (default `--keep-dup 1` destroys signal at high-coverage transcripts).

## Step 6: DRACH Motif Sanity Check

```bash
findMotifsGenome.pl \
    exomePeak2_output/Mod.bed \
    hg38 motif_output \
    -rna -size 100 -len 5,6 -p 8
```

Report DRACH enrichment on the peak set as a sanity check (P-value < 1e-50 expected). NEVER post-hoc filter individual peaks by DRACH.

## Step 7: exomePeak2 Differential (Control vs Treatment)

**Goal:** Identify m6A peaks that change between control and treatment conditions, with per-peak log2FC + FDR, using exomePeak2's integrated peak-calling + differential interface.

**Approach:** Populate `bam_ip` + `bam_input` with the control arm and `bam_treated_ip` + `bam_treated_input` with the treatment arm; populating the treated arms triggers differential mode (there is NO `mode=` argument). Apply effect-size + FDR filters downstream.

```r
library(exomePeak2)
library(GenomicFeatures)
library(BSgenome.Hsapiens.UCSC.hg38)

txdb <- makeTxDbFromGFF('refs/annotation.gtf', format='gtf')

ctrl_ip     <- c('aligned/ctrl_IP1.bam', 'aligned/ctrl_IP2.bam', 'aligned/ctrl_IP3.bam')
ctrl_input  <- c('aligned/ctrl_Input1.bam', 'aligned/ctrl_Input2.bam', 'aligned/ctrl_Input3.bam')
treat_ip    <- c('aligned/treat_IP1.bam', 'aligned/treat_IP2.bam', 'aligned/treat_IP3.bam')
treat_input <- c('aligned/treat_Input1.bam', 'aligned/treat_Input2.bam', 'aligned/treat_Input3.bam')

diff_result <- exomePeak2(
    bam_ip            = ctrl_ip,
    bam_input         = ctrl_input,
    bam_treated_ip    = treat_ip,
    bam_treated_input = treat_input,
    txdb              = txdb,
    bsgenome          = BSgenome.Hsapiens.UCSC.hg38,   # bsgenome=, not genome=
    paired_end        = TRUE,
    library_type      = 'unstranded',
    peak_calling_mode = 'exon',
    save_dir          = 'exomePeak2_diff_output'       # writes DiffMod.bed / DiffMod.csv; no experiment_name arg
)

diff_table <- Results(diff_result)   # SummarizedExomePeak has no as.data.frame method; Results() returns the data.frame
# differential effect-size column is DiffModLog2FC (not log2FC)
sig <- diff_table[diff_table$padj < 0.05 & abs(diff_table$DiffModLog2FC) > 0.5, ]
nrow(sig)
```

exomePeak2 has NO `mode=` or `experiment_design=` argument. Populating `bam_treated_ip` + `bam_treated_input` triggers differential output. For batch / antibody-lot covariate adjustment, fall through to featureCounts-on-peaks -> DESeq2 (see `epitranscriptomics/m6a-differential`).

## Step 8: Peak Annotation to Transcript Features

```r
library(ChIPseeker)
library(TxDb.Hsapiens.UCSC.hg38.knownGene)
library(rtracklayer)

peaks <- import('exomePeak2_output/Mod.bed')
anno <- annotatePeak(peaks, TxDb=TxDb.Hsapiens.UCSC.hg38.knownGene, level='transcript')
plotAnnoBar(anno)
plotDistToTSS(anno)
```

Flag peaks within ~50 nt of TSS as m6A-or-m6Am ambiguous (antibody cross-reactivity with PCIF1-deposited cap m6Am).

## Step 9: Guitar Metagene (Biological QC Anchor)

```r
library(Guitar)
library(TxDb.Hsapiens.UCSC.hg38.knownGene)

GuitarPlot(
    txTxdb          = TxDb.Hsapiens.UCSC.hg38.knownGene,
    stBedFiles      = list('exomePeak2_output/Mod.bed'),
    miscOutFilePrefix = 'figures/m6a_metagene'
)
```

Expected pattern: peak density rises toward and peaks near the stop codon (3'UTR-proximal end of CDS). If absent, suspect IP failure or wrong antibody; do NOT proceed to downstream interpretation.

## Complete Bash Driver

```bash
#!/usr/bin/env bash
set -euo pipefail

STAR_INDEX=$1
GTF=$2
IP_R1=$3
IP_R2=$4
INPUT_R1=$5
INPUT_R2=$6
OUTPUT_DIR=$7

mkdir -p "${OUTPUT_DIR}"/{qc,trimmed,aligned,peaks,figures}

fastp --in1 "${IP_R1}" --in2 "${IP_R2}" \
    --out1 "${OUTPUT_DIR}/trimmed/IP_R1.fq.gz" --out2 "${OUTPUT_DIR}/trimmed/IP_R2.fq.gz" \
    --json "${OUTPUT_DIR}/qc/IP_fastp.json" --length_required 25 --detect_adapter_for_pe --thread 8

fastp --in1 "${INPUT_R1}" --in2 "${INPUT_R2}" \
    --out1 "${OUTPUT_DIR}/trimmed/Input_R1.fq.gz" --out2 "${OUTPUT_DIR}/trimmed/Input_R2.fq.gz" \
    --json "${OUTPUT_DIR}/qc/Input_fastp.json" --length_required 25 --detect_adapter_for_pe --thread 8

for sample in IP Input; do
    STAR --runMode alignReads --genomeDir "${STAR_INDEX}" \
        --readFilesIn "${OUTPUT_DIR}/trimmed/${sample}_R1.fq.gz" "${OUTPUT_DIR}/trimmed/${sample}_R2.fq.gz" \
        --readFilesCommand zcat --outSAMtype BAM SortedByCoordinate \
        --outFilterMultimapNmax 20 \
        --outFileNamePrefix "${OUTPUT_DIR}/aligned/${sample}_" --runThreadN 12
    samtools index "${OUTPUT_DIR}/aligned/${sample}_Aligned.sortedByCoord.out.bam"
done

macs3 callpeak \
    --treatment "${OUTPUT_DIR}/aligned/IP_Aligned.sortedByCoord.out.bam" \
    --control "${OUTPUT_DIR}/aligned/Input_Aligned.sortedByCoord.out.bam" \
    --format BAMPE --gsize hs --nomodel --extsize 150 --keep-dup all \
    --broad --broad-cutoff 0.1 --qvalue 0.05 \
    --outdir "${OUTPUT_DIR}/peaks" --name m6a
```

The full pipeline (incl. exomePeak2 peak calling, DRACH check, ChIPseeker annotation, Guitar metagene) is best orchestrated in Snakemake or Nextflow with the per-skill recipes from the four `epitranscriptomics/` skills.

## QC Checkpoints

| Checkpoint | Expected | Action if Failed |
|------------|----------|------------------|
| Properly-paired rate (samtools flagstat) | >=85% | Check trimming and adapter contamination |
| Replicate Spearman within condition (10 kb bins) | >=0.85 IP-IP | Inspect divergent replicate; consider exclusion |
| plotFingerprint IP-vs-input JS distance | >=0.5 | Suspect failed IP if lower |
| Saturation plateau depth | ~30-60M unique reads | Sequence deeper if not plateaued |
| DRACH motif enrichment (HOMER, peak set) | P-value < 1e-50 | Suspect IP failure or wrong antibody |
| Stop-codon enrichment in Guitar metagene | Clear 3'UTR-proximal peak | Suspect IP failure, wrong antibody, or non-m6A modification |
| 5'UTR peaks fraction | Note ambiguity zone (~50 nt of TSS) | Flag as m6A-or-m6Am ambiguous; PCIF1 cross-reactivity |

## Output Files

| File | Description |
|------|-------------|
| `exomePeak2_output/Mod.bed` | exomePeak2 peak BED12 |
| `exomePeak2_diff_output/DiffMod.bed` | Differential peaks with log2FC + FDR |
| `motif_output/` | HOMER DRACH motif enrichment report |
| `figures/m6a_metagene.pdf` | Guitar transcript-feature metagene (stop-codon QC anchor) |
| `qc/replicate_correlation.pdf` | deepTools Spearman heatmap |
| `qc/fingerprint.pdf` | deepTools Lorenz IP-enrichment plot |
| `qc/IP_rep*_lc_extrap.txt` | PreSeq saturation curves |

## Common Errors

| Symptom | Cause | Fix |
|---------|-------|-----|
| Strongest m6A peaks (high-expression transcripts) vanish | Deduplicated non-UMI MeRIP, or MACS3 default `--keep-dup 1` | Keep all reads (`--keep-dup all`); never dedup non-UMI MeRIP |
| exomePeak2 runs but gives no differential output | Expected a `mode=`/`experiment_design=` argument | Populate `bam_treated_ip` + `bam_treated_input` to trigger differential |
| Real non-canonical m6A sites lost | Filtered individual peaks by DRACH presence | DRACH is a peak-SET sanity check (E<1e-50), never a per-peak filter |
| Peak counts "differ" between conditions but it's depth | Compared raw peak numbers at unequal depth | Rarefy BAMs to a common unique-read depth before cross-condition counts |
| No stop-codon enrichment in the metagene | IP failure, wrong antibody, or non-m6A signal | STOP; do not interpret downstream (Guitar go/no-go) |
| 5'UTR peaks over-interpreted as m6A | Antibody cross-reacts with cap-adjacent m6Am (PCIF1) | Flag peaks within ~50 nt of TSS as m6A-or-m6Am ambiguous |

## References

- Dominissini D, Moshitch-Moshkovitz S, Schwartz S, et al (2012) Topology of the human and mouse m6A RNA methylomes revealed by m6A-seq. *Nature* 485:201-206. DOI 10.1038/nature11112. (MeRIP/m6A-seq; stop-codon enrichment.)
- Meyer KD, Saletore Y, Zumbo P, et al (2012) Comprehensive analysis of mRNA methylation reveals enrichment in 3' UTRs and near stop codons. *Cell* 149:1635-1646. DOI 10.1016/j.cell.2012.05.003.
- Meng J, Lu Z, Liu H, et al (2014) A protocol for RNA methylation differential analysis with MeRIP-Seq data and the exomePeak R/Bioconductor package. *Methods* 69:274-281. DOI 10.1016/j.ymeth.2014.06.008. (exome-based peak calling.)
- Cui X, Wei Z, Zhang L, et al (2016) Guitar: an R/Bioconductor package for gene annotation guided transcriptomic analysis of RNA-related genomic features. *BioMed Research International* 2016:8367534. DOI 10.1155/2016/8367534. (transcript-feature metagene.)

## Related Skills

- epitranscriptomics/merip-preprocessing - Per-step preprocessing (trim, align, QC, saturation, IP-over-Input bigWig)
- epitranscriptomics/m6a-peak-calling - exomePeak2 / MeTPeak / MACS3 deep treatment, DRACH sanity check, m6A-vs-m6Am 5'UTR flag
- epitranscriptomics/m6a-differential - Differential methods (exomePeak2, QNB, RADAR), batch / lot covariate handling, stoichiometry-vs-expression confound
- epitranscriptomics/modification-visualization - Guitar metagene, peak-centred heatmaps, pyGenomeTracks browser figures
- epitranscriptomics/m6anet-analysis - ONT direct-RNA alternative for orthogonal stoichiometry validation
- chip-seq/peak-calling - Sibling IP-vs-input peak-calling framework
- chip-seq/chipseq-qc - IP enrichment QC concepts that transfer to MeRIP
- read-alignment/star-alignment - General STAR splice-aware alignment
- workflow-management/snakemake-workflows - Snakemake orchestration patterns
- workflow-management/nextflow-pipelines - Nextflow orchestration patterns
- workflows/rnaseq-to-de - General RNA-seq -> DE pipeline patterns