Chromatin Profiling Peak Calling: bowtie2 + MACS3, as a Snakemake DAG

Author: Syed Nurul Hasan, Ph.D. Contact: nayanchoton1789@gmail.com | ORCID: 0000-0002-4564-872X

---

Why this project

CUT&RUN was named specifically in several role-tailored CVs, with no public repo behind it. This one is a real bowtie2 → peak-calling pipeline on real public data, run as a Snakemake DAG with a SLURM profile — the same infrastructure pattern as my bulk-rnaseq-star-deseq2 repo.

Important disclosure, up front: the public dataset used here (GSE145187, Kaya-Okur et al., Nat Protoc 2020) is CUT&Tag, not CUT&RUN. I'm naming that plainly rather than letting the similarity paper over it. CUT&Tag and CUT&RUN are both Henikoff-lab in-situ chromatin profiling methods (antibody targeting → controlled cleavage → sequencing) with the same downstream bioinformatics: paired-end alignment with permissive/local settings, IgG- control peak calling, TSS-proximity QC. nf-core's own cutandrun pipeline uses this exact GSE145187 dataset as its CI test fixture for that reason. This repo demonstrates that shared analysis pipeline; it is not a substitute for having actually run CUT&RUN-specific wet-lab or library-prep work.

Pipeline

Paired-end FASTQ (H3K4me3 x2 reps, IgG control x2 reps, 10,000 reads each)
        │
        ▼
1. bowtie2_index / bowtie2_align  (Docker: biocontainers/bowtie2)
   --local --very-sensitive-local --no-mixed --no-discordant -I 10 -X 700
   (standard CUT&RUN/CUT&Tag alignment settings -- short local alignments,
   permissive fragment size range)
        │
        ▼
2. sort_index_bam  (samtools)
        │
        ▼
3. macs3_callpeak  (Docker: biocontainers/macs3)
   H3K4me3 vs. matched IgG control, per replicate, BAMPE mode
        │
        ▼
4. tss_enrichment  (scripts/tss_enrichment.py)
   fraction of peak summits within 2kb of a chr20 gene TSS

Dataset

FASTQ subsets (downsampled to 10,000 reads/sample) + a chr20 reference from nf-core/test-datasets (cutandrun branch), derived from GSE145187 — Kaya-Okur HS, Janssens DH, Henikoff JG, et al. "Efficient low-cost chromatin profiling with CUT&Tag." Nat Protoc. 2020;15(10):3264-3283. H3K4me3 marks active promoters; IgG is the isotype-control background for peak calling.

Why this alignment rate is low, and that's expected: these FASTQs are reads subsampled genome-wide (not restricted to chr20), so only ~2–3% of the 10,000 reads per sample land on chr20 (~250-300 fragments after alignment/dedup). That's exactly what nf-core's own CI test fixture expects for this pairing — it verifies pipeline mechanics on a fast, tiny dataset, not statistical power. See "Results" below for why the signal is still real despite the small numbers.

MACS2 → MACS3 substitution, also disclosed: the biocontainers MACS2 image has a binary-compatibility bug under x86_64 emulation on this Apple Silicon Mac (undefined symbol: __log_finite — a glibc/libm mismatch in the compiled Cython extension). MACS3 is the actively maintained successor with the same callpeak interface and algorithm; switching to it was a same-day fix, not a workaround of anything algorithmic.

Results (this run)

• Alignment: 2.51–3.08% overall rate across all 4 samples (expected — see above).
• MACS3 (-q 0.1, BAMPE): 82 peaks (rep1), 93 peaks (rep2).
• TSS enrichment sanity check: 90.2% of rep1 summits and 65.6% of rep2 summits fall within 2kb of a chr20 gene TSS (median distance 434bp and 583bp) — the expected promoter-proximal enrichment pattern for H3K4me3, recovered cleanly even from a 10,000-read subsample. Full table: results/tss_enrichment.tsv.

Running it

bash scripts/download_data.sh

python3 -m venv .venv && source .venv/bin/activate
pip install snakemake

snakemake -s workflow/Snakefile --cores 4 -p

# SLURM cluster execution (real profile, not run here -- no cluster in this environment)
snakemake -s workflow/Snakefile --workflow-profile workflow/profiles/slurm

Key Features

• Real bowtie2 alignment + MACS3 peak calling, containerized
• IgG-controlled peak calling per replicate, matching the standard CUT&RUN/CUT&Tag analysis convention
• An actual Snakemake DAG with a SLURM executor profile
• A quantitative biological sanity check (TSS enrichment), not just "it ran"
• Both real limitations in this run (low alignment %, MACS2→3 swap, CUT&Tag vs. CUT&RUN) disclosed rather than smoothed over

References

1. Kaya-Okur HS, Janssens DH, Henikoff JG, et al. Efficient low-cost chromatin profiling with CUT&Tag. Nat Protoc. 2020;15(10):3264-3283.
2. Skene PJ, Henikoff S. An efficient targeted nuclease strategy for high-resolution mapping of DNA binding sites. eLife. 2017;6:e21856. (CUT&RUN)
3. Zhang Y, Liu T, Meyer CA, et al. Model-based Analysis of ChIP-Seq (MACS). Genome Biology. 2008;9:R137.

License

MIT — see LICENSE.