Public bisulfite & EM-seq data over the CD55 window

Where to find published CpG methylation for the same 6,500 bp / 137-CpG region we sequence with nanopore — across T cells, K562 and dozens of other sorted human cell types. Every entry links to the primary repository, with the exact coordinates and the commands to cut the window out.

Region of interest

GRCh38 / hg38 chr1:207318058-207324558

What this page is — and is not. This is a verified index of where the public data live, not a re-analysis of it. No methylation values are plotted here, because none of these datasets have been downloaded and processed for this window yet. Everything below was checked against the primary repository. Section 4 renders real values as soon as the extractor script has been run and its JSON committed — until then it stays empty on purpose.

1 · Coordinates, and the assembly trap

The window is the same 6,500 bp / 137 CpGs used throughout this site. Most public methylomes are aligned to GRCh38 or, for anything older than about 2016, to GRCh37/hg19 — so the assembly is the first thing to get right.

T2T-CHM13v2.0 hs1 chr1:206583354-206589854 Our nanopore reference. UCSC calls this assembly hs1.

GRCh38 / hg38 hg38 chr1:207318058-207324558 The lingua franca of public methylomes — ENCODE, the methylation atlas, BLUEPRINT reprocessing.

GRCh37 / hg19 liftOver Not listed here on purpose. Roadmap Epigenomics, TCGA 450K and most pre-2016 GEO submissions are hg19; convert the hg38 window yourself rather than trusting a number typed from memory.

Watch the off-by-one: BED and bigWig are 0-based half-open, while the coordinates above are written in the 1-based inclusive style the genome browsers use. A CpG is two bases, so also decide up front whether you are reporting the C on the plus strand, the G on the minus strand, or the strand-collapsed pair.

2 · Open the window in a browser

The fastest look before downloading anything. Each link jumps straight to the region; the first one also loads the Human Methylation Atlas track hub, which puts sorted CD4 T, CD8 T, naive T, B, NK and monocyte methylomes on screen at once.

UCSC hg38 + Methylation Atlas hub 39 sorted human cell types, WGBS, at our window. If the hub does not attach, add hub.txt manually under My Data → Track Hubs. UCSC T2T (hs1) The window in our own reference — useful for checking gene models and repeats against the nanopore coordinates. ENCODE · all WGBS experiments Filter by biosample on the left to reach K562, T cells and primary tissues. Each experiment page carries per-CpG bed and bigWig files. UCSC liftOver Convert the hg38 window to hg19 before touching Roadmap or any older GEO submission.

3 · Dataset catalogue

Whole-genome assays (WGBS, EM-seq) cover this window by construction; array and capture assays may not have a single probe here, so those rows are flagged. Filter by what you need.

—

Dataset	Cell type	Assay	Assembly	Access

4 · Loaded tracks

Once fetch_public_cd55_roi.py has run, its output lands in docs/data/public_cd55_roi.json and every dataset below is drawn straight from it — one mark per CpG, positioned by base pair across the window.

5 · Cutting the window out

All snippets use the hg38 window. Swap in the T2T coordinates only for files that were actually aligned to T2T-CHM13v2.0 — and check the contig naming in the header first (chr1 vs NC_060925.1).

From a bigWig (ENCODE, Roadmap, BLUEPRINT)

# hg38 window
CHR=chr1; START=207318058; END=207324558

# methylation fraction and read depth come as two separate bigWigs
bigWigToBedGraph -chrom=$CHR -start=$START -end=$END frac.bigWig  roi.frac.bedGraph
bigWigToBedGraph -chrom=$CHR -start=$START -end=$END cov.bigWig   roi.cov.bedGraph

# join on position; drop CpGs below your depth cutoff before averaging
join -j2 <(sort -k2,2 roi.frac.bedGraph) <(sort -k2,2 roi.cov.bedGraph) > roi.joined.txt

From a per-CpG BED (ENCODE bedMethyl-style)

# ENCODE WGBS experiments ship a per-CpG bed; sort, compress, index once
sort -k1,1 -k2,2n ENCFFxxxxxx.bed | bgzip > cpg.bed.gz
tabix -p bed cpg.bed.gz
tabix cpg.bed.gz chr1:207318058-207324558 > roi.cpg.bed

From raw EM-seq / WGBS reads (SRA)

# e.g. the matched NEB EM-seq vs WGBS pair on NA12878
fasterq-dump --split-files SRR10532145
bwameth.py --reference GRCh38.fa SRR10532145_1.fastq SRR10532145_2.fastq \
  | samtools sort -o em.bam -
samtools index em.bam

# per-CpG calls restricted to the window
MethylDackel extract -r chr1:207318058-207324558 --mergeContext GRCh38.fa em.bam

From the Human Methylation Atlas (wgbstools .beta files)

# the atlas is distributed as .beta files indexed by CpG number, not by bp
wgbstools convert -r chr1:207318058-207324558     # → CpG index range for the window
wgbstools beta_to_table -r chr1:207318058-207324558 *.beta > roi_atlas.tsv

check flag names against `wgbstools --help`; the CLI has changed between releases

Whatever the source, keep the per-CpG depth alongside the fraction. A public methylome at 5× over this window gives a fraction that looks precise and is not — the same trap as reading a single nanopore molecule as if it were a population average.

6 · What each platform actually measures

These are not three ways of measuring the same thing, and the differences matter when a public track is used to sanity-check our nanopore calls.

Platform	Reads out	Watch for
WGBS	5mC + 5hmC, indistinguishable	Bisulfite fragments DNA and biases coverage; unconverted C reads as methylated. Older datasets often sit at low depth.
EM-seq	5mC + 5hmC, indistinguishable	Enzymatic, so less DNA damage and more even coverage than WGBS at equal depth — but the same 5mC/5hmC ambiguity unless a separate protocol is run.
Nanopore	5mC and 5hmC separately, per molecule	Our own data. Per-read calls make single-molecule patterns visible, at the cost of a model-dependent call with its own error profile.
450K / EPIC	Beta value at fixed probe positions	Only a few CpGs per locus, if any. Check the manifest for probes inside the window before assuming coverage.

7 · Before comparing any of this to our data

Public K562 is unsorted. ENCODE K562 methylomes are a mixed population, not the CD55-LOW / CD55-HIGH sorted fractions. They give a baseline for the locus, not a comparator for the sorted groups.
Public T cells are unedited and unstimulated (mostly). They are a reference for what the CD55 promoter looks like without an editor, which is exactly the point — but donor, subset and activation state all move methylation.
Bulk versus single-molecule. A bigWig fraction is a population average. Our nanopore data carry per-molecule patterns; collapsing ours to a fraction to compare throws away the part that is ours.
Depth is not uniform. Genome-wide assays cover this window, but not always deeply. Carry the coverage track and set a threshold before drawing anything.
Colour convention holds here too. Methylated / silenced stays orange, unmethylated / active stays blue, and no chart on this site encodes a group by colour alone.

8 · Sources

ENCODE portal — encodeproject.org
ENCODE K562 WGBS (HAIB / Myers) — ENCSR765JPC
ENCODE T-cell WGBS (Roadmap-submitted, adult CD3) — ENCSR663MXB
Loyfer et al. 2023, Nature, DNA methylation atlas of normal human cell types — PMC9811898
Methylation Atlas UCSC hub — Meth_Atlas_hg38/hub.txt; tooling: wgbs_tools
Vaisvila et al. 2021, Genome Research, EM-seq — gr.266551.120; pipeline: nebiolabs/EM-seq
Roadmap Epigenomics processed data — egg2.wustl.edu
BLUEPRINT sorted blood WGBS, Bioconductor wrapper — FlowSorted.Blood.WGBS.Blueprint
Durek et al. 2016, Immunity, CD4 memory T cell epigenomes (DEEP) — PMID 27851915
MEXPRESS, TCGA methylation by gene — mexpress.ugent.be
UCSC Genome Browser REST API — api docs