Annotation for Infinium DNA Methylation BeadChips

Probe-level annotation for the Illumina Infinium DNA methylation arrays: genomic coordinates, mapping quality, design/quality masks, and KnowYourCG feature sets (chromatin states, transcription-factor binding, CpG islands, imprinting, …). Every per-probe file is pinned positionally to one canonical probe ordering per platform, and each release is an internally coherent, SHA256SUMS-verified, git-tagged snapshot.

• Interactive browser & downloads: https://zwdzwd.github.io/InfiniumAnnotation
• Analyze arrays in R: sesame (Bioconductor)
• From IDATs on the command line: sesame-cli
• Feature enrichment: KnowYourCG
• Read/query the .cm files: YAME

Platforms

Platform	Array	Manifest	Genome	Probes	Reference
MSA	Infinium Methylation Screening Array (MSA-48v1-0)	B1	hg38	284,309	Goldberg 2025
EPICv2	Infinium MethylationEPIC v2.0 (EPIC-8v2-0)	A2	hg38	937,690	Kaur 2024
EPIC	Infinium MethylationEPIC v1.0	B5	hg38	866,553	Zhou 2016
HM450	HumanMethylation450	v1.2	hg38	486,427	Zhou 2016
HM27	HumanMethylation27	v1.2	hg38	27,722	Zhou 2016
MM285	Infinium Mouse Methylation (MouseMethylation-12v1-0)	A2	mm10, mm39	287,692	Zhou 2022
Mammal40	HorvathMammal40 Mammalian Methylation	Canonical 3.2019	hg38	38,607	Arneson 2022

Probe counts are exact rows in each <PLAT>.ordering.tsv.gz. Each manifest revision is the current/final one Illumina ships for that array.

What's in each platform directory

Files are row-aligned: data row i of every file below is the same probe. The ordering is the only file that carries probe names — it is the index.

File	Contents
<PLAT>.ordering.tsv.gz	The probe index. Probe_ID, M/U (bead addresses), col (color channel). Genome-independent; derived from the manufacturer manifest, LC_ALL=C sorted.
<PLAT>.<genome>.coord.tsv.gz	Per-probe genomic coordinate: CpG_chrm, CpG_beg (0-based), strand, mapQ. Positional; no probe IDs stored (see below).
<PLAT>.<genome>.mask.cm (+.idx)	Design/quality masks (M_mapping, M_nonuniq, SNP masks, …) as a YAME .cm bitset — one record per mask. Positional.
<PLAT>.typeI_ext.tsv.gz	Infinium-I extension base for the GCT bisulfite-conversion control: C (extension C — Grn here means unconverted), T (background reference), . otherwise. M_general-masked probes are excluded. Positional.
KYCG/	KnowYourCG feature sets as .cm (one dir, its own SHA256SUMS). Positional.
SHA256SUMS	Binds the set. Verify with cd <PLAT> && sha256sum -c SHA256SUMS.

strand is the interrogated cytosine's strand (+ = the C of the CpG on the top strand at CpG_beg; - = the C complementing the G) — it matches sesame's probe_strand. The interval width is implied by probe class (from the Probe_ID prefix): a cg probe is a 2-bp CpG (end = CpG_beg + 2); ch/rs/nv probes target a single base (end = CpG_beg + 1).

Matching probe names to the files

The coord, mask.cm, and KYCG/ files are positional to the ordering and do not store probe IDs (the IDs live once, in the ordering). Attach them with sesame attach-probe, which prepends the ordering's Probe_ID to any positional file — a coord .tsv.gz, a .cm mask/feature, or a preprocess beta .cg:

# Coordinate table with probe names:
sesame attach-probe --index MSA.ordering.tsv.gz MSA.hg38.coord.tsv.gz
#   Probe_ID   CpG_chrm   CpG_beg   strand   mapQ

# Mask matrix with probe names (unpacks the .cm and prepends Probe_ID):
sesame attach-probe --index MSA.ordering.tsv.gz MSA.hg38.mask.cm
#   Probe_ID   <one column per mask tag>

Use --platform MSA instead of --index to pull the ordering from the fetched store. Without the binary, the same pairing is a positional paste, e.g. paste <(zcat MSA.ordering.tsv.gz | cut -f1) <(zcat MSA.hg38.coord.tsv.gz).

• sesame-cli — attach-probe (above); also preprocess (IDATs → beta .cg in this same ordering), dml, cnv.
• YAME reads/queries the .cm files directly (yame unpack, yame summary, …); rows come out in ordering order.
• In R, sesame and KnowYourCG consume these directly.

Versions

The repository is released as git tags; each tag is a self-consistent snapshot (every file bound by SHA256SUMS). Use the latest tag (v7) unless you need to reproduce an older result.

Common build parameters across all platforms:

• Mapping: BISCUIT (bisulfite-aware BWA-MEM) with an added primary-chromosome preference pass (biscuit postalt --unplaced) that keeps a probe on the main assembly when it also hits an equal-scoring unplaced/_alt contig.
• Ordering: from the manufacturer manifest (bead addresses + color channel), LC_ALL=C sorted — reproducible from any shell, authoritative for addresses.
• coord strand: the interrogated-cytosine strand (SAM FLAG), verified 100% against sesame's probe_strand.
• Masks: mapping masks (M_mapping, M_nonuniq) from the alignment; SNP masks from dbSNP (human) / Mouse Genomes Project strain SNPs (mm10).

Tag history: v1 MSA · v2 +postalt, +EPICv2 · v3 +HM450/EPIC, +coord · v4 coord positional layout · v5 coord strand fix · v6 +MM285 (mm10+mm39) · v7 +Mammal40/HM27.

References

Per platform:

• MSA — Goldberg et al., Scalable screening of ternary-code DNA methylation dynamics associated with human traits, Cell Genomics 2025.
• EPICv2 — Kaur et al., Comprehensive evaluation of the Infinium human MethylationEPIC v2 BeadChip, Epigenetics Communications 2024.
• EPIC / HM450 / HM27 — Zhou, Laird & Shen, Comprehensive characterization, annotation and innovative use of Infinium DNA methylation BeadChip probes, Nucleic Acids Research 2016.
• MM285 — Zhou, Hinoue, Barnes et al., DNA methylation dynamics and dysregulation delineated by high-throughput profiling in the mouse, Cell Genomics 2022.
• Mammal40 — Arneson, Haghani et al. (Horvath), A mammalian methylation array for profiling methylation levels at conserved sequences, Nature Communications 2022.

Methods:

• Feature interpretation (KnowYourCG) — Goldberg*, Fu*, et al., KnowYourCG: Facilitating base-level sparse methylome interpretation, Science Advances 2025.