← Infinium annotation

MM285 probe design groups

How each design group on the Infinium Mouse Methylation BeadChip (MM285) was selected: the source data, the selection criteria and how many probes carry the label.
The design fileDesignability Genomic featuresTarget biology Germline & early developmentRandom, CpH, SNP & technical Known issuesReferences

The design file

MM285.design.tsv.gz has two columns, Probe_ID and design, one row per probe on the 296,070-probe design-paper manifest. The design field lists every group the probe was selected for, separated by commas. A probe picked by two strategies carries both, for example CTCF,EPIC;cg04056615. Within a group entry, semicolons separate the group name from its details, such as the syntenic human EPIC probe or the nearest TSS. The last column of each table below shows that format.

The same groups are distributed as the knowYourCG (v1) knowledgebase KYCG.MM285.designGroup.20210210 (ExperimentHub EH6825, getDBs("MM285.designGroup")). The design and the selection criteria are described in Zhou et al. 2022 [1]. This page adds the criteria from the original selection code where the paper is silent, and notes where the two differ.

Counts. The "Probes" column counts probes carrying each label in the design file. A probe can carry several labels, so the column does not sum to the array size. The paper counts label occurrences instead, so a probe near two TSSs counts twice there. That is why, for example, the paper gives 100,948 TSS CpGs against 69,237 probes here.

Designability: the shared pre-filter

Every CpG group draws from one list of designable CpGs [1]. For each CpG in mm10, both Infinium-I probe designs on the converted strand were aligned with BISCUIT. A design was kept only if its 30-, 35- and 40-nt 3′ subsequences and the full probe all map uniquely, with mapping quality above 20 for both alleles. Designs were then dropped if a SNP from 15 laboratory strains [2] or another CpG lay within 10 nt of the 3′ end, or if the probe held six or more CpGs. A design score for GC content and melting temperature had to exceed 0.3. About 9.77 million CpGs passed. Some groups draw from a stricter high-quality subset.

Genomic features

GroupProbesSourceSelectionDetail format
TSS69,237Ensembl release 96 protein-coding transcripts, collapsed to 54,376 TSSs [3] Designable CpGs within 1,500 bp of a TSS, ranked by distance. Lower-quality designs (score under 0.6) carry a 500 bp penalty. At most two CpGs per TSS. 1,972 TSSs (3.6%) have no designable CpG. TSS;chr:pos;gene;transcript;distance;rank
lincRNATSS12,029Ensembl 96 lincRNA TSSs (8,621) [3] At most two CpGs per TSS, closest first. 1,327 lincRNA TSSs have no designable CpG. lincRNATSS;chr:pos;gene;transcript;distance
PseudogeneTSS10,244Ensembl 96 pseudogene TSSs [3] Closest designable CpG to each TSS.PseudogeneTSS;chr:pos;gene;transcript;distance
miRNA3,970Ensembl 96 miRNA transcripts [3] Closest designable CpGs to each transcript start.miRNA;chr:pos;gene;transcript;distance;rank
CGI17,134UCSC mm10 CpG islands One CpG for each island at an Ensembl TSS, three for each island away from any TSS. Covers 13,365 of 16,023 islands.CGI;chr:beg-end
GeneBody24,638RefSeq canonical protein-coding transcripts [4] One random CpG per transcript, from 2 kb downstream of the TSS to the transcription end, excluding 5′ CpG islands. Transcripts shorter than 2 kb were skipped.GeneBody;gene;RefSeq;location;splice=0|1
Enhancer58,759ChromHMM enhancer states (strong and poised) in mouse ESCs [5] and adult tissues [6]; 176 public WGBS methylomes (paper Table S1) High-quality CpGs in an enhancer state and over 1 kb from a protein-coding TSS. Kept if methylated (>0.7) in at least five methylomes and unmethylated (<0.3) in at least five, which left 757,439 tissue-variable CpGs. 30,000 were sampled from CpGs in three or more enhancer calls and 30,000 from the rest.Enhancer
EnhancerVista1,245VISTA Enhancer Browser, experimentally validated enhancers [7] High-quality CpGs inside validated mouse enhancer elements.EnhancerVista;chr:beg-end
CTCF8,616Mouse ENCODE CTCF ChIP-seq, 19 tissues [8], checked against ChromHMM [6] CpGs inside CTCF peaks in at least two tissues. They were binned by the fraction of compendium methylomes in which they are unmethylated (<0.2), and up to 1,000 were sampled per bin. Sites methylated in every sample were dropped.CTCF
EPIC29,008Human MethylationEPIC CpGs [9], UCSC liftOver to mm10 EPIC CpGs that lift over to a designable mouse CpG. 28,719 map uniquely in both genomes. The mouse probe sequence differs from the human one. The set is promoter-rich because EPIC is.EPIC;<EPIC probe ID>

Target biology

GroupProbesSourceSelectionDetail format
MonoallelicMeth7,813WGBS compendium, excluding germ-cell and early-embryo samples Autosomal CpGs measured in more than six samples and intermediately methylated (0.35 to 0.65) in over 65% of them. Intended to capture imprinting and other mono-allelic methylation.MonoallelicMeth
ImprintDMR62326 curated imprinted regions (paper key resources table) High-quality CpGs inside known imprinting control regions and secondary DMRs.ImprintDMR;chr:beg-end;allele;DMR name
PlacentaIntermed981 1 sampleOne placenta WGBS sample, Hon et al. 2013, GSM1051161 [10] Intended for placenta-specific imprinting. CpGs between 0.4 and 0.6 in that sample, about 5.8 million, of which 1,000 were sampled. No comparison with other tissues was made, so the check is the absolute level. It holds in aggregate: the group median is 0.54 across 16 MM285 placentas, against 0.82 in somatic tissue. Per-probe agreement has not been measured.PlacentaIntermed
VMR5,835Variably methylated IAP elements, Kazachenka et al. 2018 [11] One CpG for each of 6,402 VM-IAP regions. 633 regions had no designable CpG.VMR;chr:beg-end
PMDsoloWCGW5,095Common PMDs and solo-WCGW CpGs [12]; lamina-associated domains [13] Random solo-WCGW CpGs, with no other CpG within 35 bp, inside common PMDs that are also lamina-associated. These CpGs lose methylation with mitotic cell division.PMDsoloWCGW
Adenoma8,311Colon tumors induced with DSS and azoxymethane, Abu-Remaileh et al. 2015 [14] CpGs methylated in tumor (>0.7) and unmethylated in normal colon (<0.3). See the note on sample numbers.Adenoma
Clock754Published mouse epigenetic clocks [15–19] All designable CpGs of six clocks: Meer 435, Petkovich 90, Stubbs 329, Thompson 273 and 582, and Wang 148. Clock;<clock>;coefficient
RMSK4,517Repbase consensus sequences [20] Internal CpGs of rodent-specific repeat families with over 1,000 genomic copies, including 242 probes for IAP and 423 for L1 families.RMSK;family|class|…;consensus position
Mitochondria32chrM Mitochondrial CpGs, expected unmethylated, for use as a technical control.Mitochondria

Germline and early development

The paper's methods describe these groups in one sentence, with no criteria [1]. The criteria below come from the original selection code. Each group compares target samples with the rest of the public WGBS compendium, after dropping methylomes with 15 million or fewer covered CpGs. That filter removed several cited studies, so some groups rest on very few samples. The tags say how many target samples were used and whether the group has been checked on the array.

GroupProbesTarget samplesSelection
PGCMeth474 5 samplesWang et al. 2014 [21]: male and female PGCs, ICM, 2-cell, 4-cell Either methylated (>0.7) in all five, or above 0.8 in both PGC samples. These CpGs resist the global demethylation of early development. Not testable on the MM285 atlas, which has no PGC or early-embryo profiles.
SpermMeth397 replicates in testisMale germline, spermatogonia to sperm, 6 samples: Hammoud et al. 2014 [22] spermatogonia (2), spermatocyte, spermatid and mature sperm; Wang 2014 sperm [21] Above 0.85 in all but one target sample, below 0.15 in over 80% of other samples, and below 0.3 in at least one zygote. The group is male-germline rather than sperm-specific, despite its name. In the 7 MM285 testis profiles, which contain that whole lineage, 99% of probes gain more than 0.2 over somatic tissue.
SpermUnmeth208 replicates in testisSame six male germline samples Below 0.15 in all but one target sample, above 0.85 in over 80% of other samples, and above 0.7 in at least one zygote. In the 7 MM285 testis profiles, 99% of probes lose more than 0.2.
OocyteMeth172 1 sampleWang 2014 oocyte [21] Above 0.85 in the oocyte and below 0.15 in over 80% of other samples. Not testable on the MM285 atlas, which has no oocyte profiles.
OocyteUnmeth486 1 sampleWang 2014 oocyte [21] Below 0.15 in the oocyte and above 0.85 in over 80% of other samples. Not testable on the MM285 atlas, which has no oocyte profiles.
ZygoteMeth474 2 samplesPeat et al. 2014 [23]: wild-type and Tet3-knockout zygote Above 0.75 in both zygotes and below 0.15 in over 80% of other samples. Not testable on the MM285 atlas, which has no zygote profiles.
ZygoteUnmeth480 2 samplesSame two zygotes Below 0.15 in both and above 0.85 in over 80% of other samples. Not testable on the MM285 atlas.
PlacentaMeth485 does not replicateOne placenta sample, Hon 2013, GSM1051161 [10] Above 0.85 in the placenta and below 0.15 in over 90% of other samples. See known issues.
PlacentaUnmeth484 does not replicateSame placenta sample Below 0.15 in the placenta and above 0.85 in over 90% of other samples. See known issues.

Random, CpH, SNP and technical probes

GroupProbesSourceSelectionDetail format
Random27,999Designability list Random designable CpGs to cover uncharacterized biology. Extra draws on chrX, and every designable chrY CpG, offset the sex chromosomes' loss to the uniqueness filter.Random
CpH2,794Adult brain WGBS, Guo et al. 2014 [24] CpA, CpC and CpT sites. Half are methylated in brain, the rest random. Same mapping and design-score filters as CpG probes.CpH;CA|CC|CT
SNP1,490Mouse Genomes Project, 36 inbred and wild strains [2] Up to ten SNPs per branch of the strain phylogeny, 591 branch-specific SNPs in all, outside repeats. Used to check genetic background.SNP;chr:pos;…;REF>ALT;strain;…
Control4,216Illumina and custom control probes Control probes, including negative and bisulfite-conversion controls.Control;type
Multi7,317— Probes intentionally designed to hit several genomic locations. SeSAMe masks them by default.Multi
UK4,541— Probes that missed the design objective through sequence mismatch or weak hybridization, with the uk prefix. SeSAMe masks them by default.UK;reason

Known issues

PlacentaMeth and PlacentaUnmeth do not replicate (2026-09). These two groups carry no placenta-specific methylation signal on the array. Do not use them as placenta markers. This affects KYCG.MM285.designGroup.20210210 (EH6825) and the same labels in MM285.design.tsv.gz.
Both rest on a single placenta methylome (GSM1051161). They do not replicate in MM285 data from 16 placentas and 682 adult somatic samples (GSE184410). PlacentaMeth probes gain no methylation in placenta: the median change is 0.00, the same as random probes with similar somatic methylation. PlacentaUnmeth probes lose less methylation than comparable highly methylated CpGs, with a median drop of 0.06 against 0.21. The labels are not swapped, since neither direction holds. Exclude both groups from any placenta-specific interpretation.
The rest of the design-group set is not affected by this finding. A corrected version of the knowledgebase is planned. The design-group set is not part of the KnowYourCG v2 (kycg) catalogue, so v2 is unaffected. Reported by a user.
Where the selection code differs from the paper.
• Placenta source. The paper cites Popp et al. 2010 [25] for the placenta groups. The code used Hon 2013 sample GSM1051161. The Popp placenta methylomes fell below the coverage filter.
• Adenoma samples. The paper describes two tumor and two normal colon samples with at least six reads. The code compared one DSS-plus-azoxymethane tumor with one normal colon, with the second pair commented out.

References

  1. Zhou W et al. DNA methylation dynamics and dysregulation delineated by high-throughput profiling in the mouse. Cell Genomics 2022.
  2. Keane TM et al. Mouse genomic variation and its effect on phenotypes and gene regulation. Nature 2011.
  3. Ensembl release 96 (2019).
  4. NCBI RefSeq.
  5. Pintacuda G et al. hnRNPK recruits PCGF3/5-PRC1 to the Xist RNA B-repeat. Mol Cell 2017.
  6. Shen Y et al. A map of the cis-regulatory sequences in the mouse genome. Nature 2012.
  7. Visel A et al. VISTA Enhancer Browser. Nucleic Acids Res 2007.
  8. Yue F et al. A comparative encyclopedia of DNA elements in the mouse genome. Nature 2014.
  9. Pidsley R et al. Critical evaluation of the Illumina MethylationEPIC BeadChip. Genome Biol 2016.
  10. Hon GC et al. Epigenetic memory at embryonic enhancers identified in DNA methylation maps from adult mouse tissues. Nat Genet 2013.
  11. Kazachenka A et al. Identification, characterization, and heritability of murine metastable epialleles. Cell 2018.
  12. Zhou W et al. DNA methylation loss in late-replicating domains is linked to mitotic cell division. Nat Genet 2018.
  13. Peric-Hupkes D et al. Molecular maps of the reorganization of genome–nuclear lamina interactions. Mol Cell 2010.
  14. Abu-Remaileh M et al. Chronic inflammation induces a novel epigenetic program that is conserved in intestinal adenomas and in colorectal cancer. Cancer Res 2015.
  15. Meer MV et al. A whole lifespan mouse multi-tissue DNA methylation clock. eLife 2018.
  16. Thompson MJ et al. A multi-tissue full lifespan epigenetic clock for mice. Aging 2018.
  17. Petkovich DA et al. Using DNA methylation profiling to evaluate biological age and longevity interventions. Cell Metab 2017.
  18. Wang T et al. Epigenetic aging signatures in mice livers are slowed by dwarfism, calorie restriction and rapamycin treatment. Genome Biol 2017.
  19. Stubbs TM et al. Multi-tissue DNA methylation age predictor in mouse. Genome Biol 2017.
  20. Bao W et al. Repbase Update. Mobile DNA 2015.
  21. Wang L et al. Programming and inheritance of parental DNA methylomes in mammals. Cell 2014.
  22. Hammoud SS et al. Chromatin and transcription transitions of mammalian adult germline stem cells and spermatogenesis. Cell Stem Cell 2014.
  23. Peat JR et al. Genome-wide bisulfite sequencing in zygotes identifies demethylation targets and maps the contribution of TET3 oxidation. Cell Rep 2014.
  24. Guo JU et al. Distribution, recognition and regulation of non-CpG methylation in the adult mammalian brain. Nat Neurosci 2014.
  25. Popp C et al. Genome-wide erasure of DNA methylation in mouse primordial germ cells is affected by AID deficiency. Nature 2010.
Questions: wanding.zhou@pennmedicine.upenn.edu