What a Methylation Lab Test Measures

Oak
methylation epigenetics dna-methylation homocysteine bioinformatics
A feathered reptile specimen on a black plinth, one side damp and cold, the other side's quills tipped with small pale beads.

Two unrelated things are sold under the name “methylation lab test.” The first is a blood panel of one-carbon metabolism: homocysteine, B12, methylmalonic acid, folate, B6 as pyridoxal-5’-phosphate, sometimes the SAM/SAH ratio, usually bundled with an MTHFR genotype. The second is a DNA methylation assay, which measures the fraction of cells in a sample carrying a methyl group at specific CpG dinucleotides, and includes everything from a single-locus clinical test like MLH1 promoter methylation to array-wide epigenetic clocks and cell-free DNA cancer screening. They share a word and nothing else. If you want to work with your own data, decide which one you are asking for before you order anything.

The biochemical panel and its preanalytic traps

The consumer “methylation panel” is a small set of standard clinical chemistry assays, and the interesting part is not the assay list but the sample handling. Homocysteine is the marker that carries most of the information, and it is also the one most often ruined in collection. Red cells continue to export homocysteine into plasma after the draw, so a tube left at room temperature for an hour can read meaningfully higher than the same blood centrifuged promptly. Ask for EDTA plasma placed on ice and spun within 30 minutes, or a tube containing a glycolysis inhibitor. If the lab cannot tell you the time from draw to spin, the number is soft.

The rest of the panel is more forgiving. Serum B12 is a poor standalone marker because much of circulating B12 is bound to haptocorrin and unavailable to cells, which is why methylmalonic acid and, where available, holotranscobalamin add real information. Red cell folate reflects a months-long average and serum folate reflects the last few meals, so the two answer different questions. Direct measurement of S-adenosylmethionine and S-adenosylhomocysteine is offered by a handful of labs, but both metabolites are unstable in whole blood and the ratio is sensitive to processing delay, so we treat a single SAM/SAH result as a research number rather than a fact about you. Interpretation of any of these, particularly an elevated homocysteine or a low B12 with high methylmalonic acid, belongs with a physician.

MTHFR genotype is already in your genome file

If you have whole-genome sequencing, you do not need to buy an MTHFR test. The two commonly reported variants are rs1801133 (C677T, p.Ala222Val) and rs1801131 (A1298C), and both are ordinary SNPs on chromosome 1. Pull them straight out of your VCF:

bcftools view -r chr1:11794419-11796321 sample.vcf.gz | \
  bcftools query -f '%CHROM\t%POS\t%ID\t%REF\t%ALT[\t%GT\t%DP\t%GQ]\n'

Check the depth and genotype quality fields before you believe the call. Below roughly 10x at the site, a heterozygote can be miscalled as homozygous reference. Note that rs1801133 is common: the T allele reaches double-digit to roughly 30–40 percent frequency across populations, and homozygous individuals are found everywhere. Professional genetics bodies have moved away from routine MTHFR genotyping precisely because the genotype alone changes little that the homocysteine measurement does not tell you better. The genotype is a fixed prior. The metabolite is the measurement.

DNA methylation assays: arrays, bisulfite sequencing, nanopore

The second meaning of the term concerns 5-methylcytosine on your DNA, overwhelmingly at CpG sites, where methylation state tracks with transcriptional repression, imprinting, X inactivation, and tissue identity 1. Three technologies dominate.

Illumina’s EPIC array (v2 covers roughly 935,000 CpGs) is the workhorse: cheap, well-standardized, and the substrate for nearly every published epigenetic clock. You receive a pair of IDAT files per sample, one per color channel. Whole-genome bisulfite sequencing covers roughly 28 million CpGs but needs about 30x coverage to give stable per-site estimates, and the bisulfite conversion chemistry fragments DNA and collapses sequence complexity, which hurts mapping. Nanopore sequencing calls 5mC directly from the raw current signal with no chemical conversion, giving you methylation and structural variants and phasing from one library. We prefer nanopore when the budget allows and arrays when it does not, because the array’s value is comparability to the published literature.

For arrays, use the sesame R package rather than minfi unless you need a legacy workflow:

betas <- openSesame("idat_dir/", prep = "QCDPB")

That preparation string masks probes overlapping common SNPs, does out-of-band background correction (noob), infers channel switching for Type I probes, and applies dye bias correction. Drop probes with detection p-value above 0.05, drop cross-reactive probes using a published masking list, and convert beta values to M-values (log2((M+1)/(U+1))) before any linear modeling, because betas are bounded at 0 and 1 and heteroscedastic.

For bisulfite sequencing, align with bwameth.py --reference hg38.fa, mark duplicates, then call with MethylDackel extract --mergeContext --minDepth 10 --OT 5,0,5,0 hg38.fa sample.bam. Spike in unmethylated lambda phage DNA and confirm conversion above 99.5 percent before trusting anything. A conversion rate of 97 percent inflates apparent methylation at every unmethylated site by about three points, which is larger than most effects you are looking for. For nanopore, basecall with dorado basecaller sup,5mCG_5hmCG, which writes MM and ML modification tags into the BAM, then summarize with modkit pileup --cpg --combine-strands --ref hg38.fa.

Cell composition is the confound that eats most blood methylation results

Whole blood is a mixture, and neutrophils, monocytes, CD4 T cells, CD8 T cells, B cells, and NK cells have grossly different methylomes. A shift in your differential count between two draws will produce hundreds of apparently significant methylation differences that have nothing to do with any regulatory change inside a cell. Always deconvolve: EpiDISH or the Houseman reference-based method in minfi will give you estimated cell fractions from the array data itself, and those fractions belong in every model as covariates. This is also why global measures such as LINE-1 retrotransposon methylation, used as a proxy for genome-wide methylation and shown to vary with environmental lead exposure, need careful covariate handling before they mean anything 2. Global blood methylation has been examined as a cancer-associated biomarker as well, with modest effect sizes 3.

Which methylation tests have clinical validation

A short list of DNA methylation assays are validated clinical tests, and they are all single-locus or small-panel cancer applications rather than wellness panels. MLH1 promoter methylation in tumor tissue distinguishes sporadic from germline causes of mismatch repair loss in colorectal cancer. Plasma SEPT9 methylation is an FDA-approved blood test for colorectal cancer screening, with real-world laboratory experience published on its performance and its false-positive burden 4. Newer multi-marker cell-free DNA methylation classifiers have been evaluated prospectively in high-risk colorectal screening cohorts 5, and multisite panels in peripheral blood mononuclear cells have been tested for early lung adenocarcinoma detection in multicenter designs 6. Systematic review of head and neck squamous cell carcinoma blood markers shows the pattern that recurs across the field: many candidate loci, few independently replicated 7. Reviews of clinical translation are frank that assay standardization and validation cohorts, not marker discovery, are the bottleneck 8.

Nothing on that list tells you whether “your methylation is working.” Epigenetic clocks (Horvath’s 353-CpG multi-tissue clock, PhenoAge’s 513 CpGs, DunedinPACE’s 173 CpGs) are computable from an EPIC array and are interesting research readouts, but single-sample test-retest reliability for the first-generation clocks is poor enough that a two-year difference between draws can be technical noise. Principal-component versions of the clocks were built specifically to fix this, and we use those. Methylation at metastable epialleles does respond to nutritional state, with season of conception in Gambian cohorts producing measurable differences in offspring 9, and mitochondrial DNA methylation has been proposed as a separate readout entirely 10. Both are live research questions rather than clinical tests.

Questions people also ask

What are signs of poor methylation? There is no validated symptom signature for it. Fatigue, brain fog, and mood changes have many causes and do not map onto one-carbon metabolism in any specific way. The measurable quantity is plasma homocysteine, supported by methylmalonic acid, B12, and folate. If yours is elevated, that is a finding to take to a physician, who will look for the common causes including B12 or folate status, kidney function, thyroid function, and medications.

Is a methylation panel the same as an MTHFR test? No. The panel measures metabolites in blood. MTHFR genotyping reads two fixed DNA variants that you can extract from an existing whole-genome VCF at no extra cost.

Should I get an EPIC array or bisulfite sequencing? Choose the array if you want results comparable to published clocks and epigenome-wide association studies, since nearly all of that literature is array-based. Choose nanopore sequencing if you want full CpG coverage, allele-specific methylation, and structural variants from one library and can handle the larger data volume.

Can a blood methylation test screen me for cancer? A few specific assays have regulatory approval or prospective evidence for defined indications, notably plasma SEPT9 for colorectal cancer 4 5. General-purpose “methylation health” panels have no such evidence, and screening decisions belong with your clinician.

Oak builds longitudinal molecular profiles of individuals: whole-genome sequencing, RNA sequencing, proteomics, blood biomarkers, and continuous glucose data, integrated into one model of you. Build your profile.

Footnotes

  1. Jörg Tost. DNA Methylation: An Introduction to the Biology and the Disease-Associated Changes of a Promising Biomarker. Molecular Biotechnology, 2009. https://doi.org/10.1007/s12033-009-9216-2 ↩

  2. Robert O. Wright, Joel Schwartz, Rosalind J. Wright, et al. Biomarkers of Lead Exposure and DNA Methylation within Retrotransposons. Environmental Health Perspectives, 2010. https://doi.org/10.1289/ehp.0901429 ↩

  3. Debra Ting Hsiung, Carmen J. Marsit, E. Andres Houseman, et al. Global DNA Methylation Level in Whole Blood as a Biomarker in Head and Neck Squamous Cell Carcinoma. Cancer Epidemiology, Biomarkers & Prevention, 2007. https://doi.org/10.1158/1055-9965.epi-06-0636 ↩

  4. Li Cai, Scott Hood, Eddie Kallam, et al. Epi proColon®: Use of a Non-Invasive SEPT9 Gene Methylation Blood Test for Colorectal Cancer Screening: A National Laboratory Experience. Journal of Clinical Epigenetics, 2018. https://doi.org/10.21767/2472-1158.100092 ↩ ↩2

  5. Fuqiang Zhao, Ping Bai, Jianfeng Xu, et al. Efficacy of cell-free DNA methylation-based blood test for colorectal cancer screening in high-risk population: a prospective cohort study. Molecular Cancer, 2023. https://doi.org/10.1186/s12943-023-01866-z ↩ ↩2

  6. Peilong Li, Shibiao Liu, Tiantian Wang, et al. Multisite DNA methylation alterations of peripheral blood mononuclear cells serve as novel biomarkers for the diagnosis of AIS/stage I lung adenocarcinoma: a multicenter cohort study. International Journal of Surgery, 2024. https://doi.org/10.1097/js9.0000000000002101 ↩

  7. Christian Sander Danstrup, Mette Marcussen, Inge Søkilde Pedersen, et al. DNA methylation biomarkers in peripheral blood of patients with head and neck squamous cell carcinomas. A systematic review. PLOS ONE, 2020. https://doi.org/10.1371/journal.pone.0244101 ↩

  8. Warwick J. Locke, Dominic Guanzon, Chenkai Ma, et al. DNA Methylation Cancer Biomarkers: Translation to the Clinic. Frontiers in Genetics, 2019. https://doi.org/10.3389/fgene.2019.01150 ↩

  9. Paula Dominguez-Salas, Sophie E. Moore, Maria S. Baker, et al. Maternal nutrition at conception modulates DNA methylation of human metastable epialleles. Nature Communications, 2014. https://doi.org/10.1038/ncomms4746 ↩

  10. Vito Iacobazzi, Alessandra Castegna, Vittoria Infantino, et al. Mitochondrial DNA methylation as a next-generation biomarker and diagnostic tool. Molecular Genetics and Metabolism, 2013. https://doi.org/10.1016/j.ymgme.2013.07.012 ↩