Genetic Methylation Testing: What It Measures and What It Cannot Tell You

Oak
methylation mthfr genomics epigenetics biomarkers
Two glass columns in a dark lab, one holding a bare blue DNA strand, the other a strand studded with glowing amber beads.

“Genetic methylation testing” refers to two unrelated measurements that got the same marketing name. The first is genotyping: reading the DNA letters at positions in genes that encode enzymes of one-carbon metabolism, most famously MTHFR c.677C>T and c.1298A>C. This tells you which protein variants you carry, and nothing about how much methylation is happening in your cells. The second is DNA methylation profiling: measuring the fraction of cells in a sample in which a given cytosine carries a methyl group, typically at hundreds of thousands to tens of millions of CpG sites. That is an epigenetic measurement, it changes with age, tissue, and exposure, and it is what “DNA methylation test” means in the research literature. If you bought a $299 kit that reported “your methylation pathway,” you bought the first one, and the reports layered on top of it are interpretation, not measurement.

Both are worth having, for different reasons, and both are more informative when you hold the underlying data rather than a PDF of colored gauges. Below is what each measurement produces, how to read it, and the specific places where the consumer interpretation layer overstates what the biology supports.

What genotyping of MTHFR and friends gives you

MTHFR encodes methylenetetrahydrofolate reductase, which converts 5,10-methylenetetrahydrofolate to 5-methyltetrahydrofolate, the methyl donor that remethylates homocysteine to methionine. The c.677C>T variant (rs1801133, p.Ala222Val) produces a thermolabile enzyme with reduced specific activity; homozygotes have roughly 30% of reference activity in vitro and modestly elevated plasma homocysteine, with the effect strongly conditioned on folate status. The c.1298A>C variant (rs1801131) has a smaller effect and is usually only discussed in compound heterozygotes. Clinical reviews of these two variants conclude that testing has narrow justification and that the genotype alone does not establish a thrombotic or obstetric risk sufficient to change management.1

If you have whole-genome sequencing, you already have these calls, plus the rest of the pathway, at better quality than an array. Pull them directly from your VCF (the variant call file, one row per position where you differ from the reference):

bcftools view -r chr1:11794419,chr1:11796321 \
  -f PASS your.vcf.gz | bcftools query -f \
  '%CHROM %POS %REF %ALT [%GT %DP %GQ]\n'

Those GRCh38 coordinates are rs1801133 and rs1801131 respectively; confirm against your reference build before trusting them, because MTHFR is on the minus strand and liftover errors between GRCh37 and GRCh38 are the single most common mistake we see in self-analysis. Check DP (read depth) is at least 20 and GQ (genotype quality) above 30. The other genes people ask about are MTR, MTRR, CBS, BHMT, SHMT1, DHFR, TCN2, and the folate transporters SLC19A1 and FOLR1. Rare loss-of-function variants in the transporters are a real subject of investigation in neural tube defect etiology, though the common polymorphisms carry small and inconsistent effects.2

The distinction that matters for your own data is between common polymorphisms and severe deficiency. Severe MTHFR deficiency is a rare inborn error caused by biallelic damaging variants, presents with homocystinuria and neurological disease, and is diagnosed by enzyme assay and metabolite measurement, not by SNP panels.3 If you are carrying two common variants and feel unwell, you do not have that condition, and the two should not be conflated the way consumer reports conflate them. Anyone with genuinely elevated homocysteine, unexplained neurological symptoms, or a thrombotic history needs a physician and a metabolic workup, not a gene report.

What genome-wide methylation profiling gives you

The direct measurement is different in kind. The Illumina MethylationEPIC array interrogates roughly 850,000 CpG sites after bisulfite conversion, returning a beta value per site between 0 and 1 that estimates the proportion of methylated alleles in the sample. It covers most annotated gene promoters and a large set of enhancer regions defined by ENCODE and FANTOM5.4 Its coverage of the CpG universe is still sparse: the human genome has about 28 million CpGs, and the array touches about 3% of them, with a known bias toward regions that were already thought to be interesting.4 Regulatory element annotation from ENCODE is what makes a differentially methylated position interpretable at all, since a beta change at an intergenic CpG with no regulatory annotation is usually uninterpretable.5

The sequencing alternative is whole-genome bisulfite sequencing (WGBS) or enzymatic methyl-seq, which covers essentially all CpGs at the cost of needing 30x or more coverage for reliable per-site estimates. For an individual profile, the practical pipeline is bwa-meth or bismark for alignment, then MethylDackel extract --mergeContext --minDepth 10 to produce a bedGraph of coverage and methylation percentage per CpG. For array data, process IDATs in R with minfi or sesame: run preprocessNoob() or the sesame pOOBAH detection-p filter, drop probes with detection p above 0.01, drop the cross-reactive and SNP-overlapping probes flagged in the published EPIC annotation, and then work with M-values (logit-transformed betas) for any statistics, because betas are heteroscedastic at the extremes.

Two interpretation traps dominate. First, blood methylation is a weighted average over neutrophils, lymphocytes, monocytes, and so on, so a shift in cell composition looks exactly like a shift in methylation. Always deconvolve with a reference-based method (FlowSorted.Blood.EPIC in minfi) and treat the estimated cell fractions as covariates before you conclude anything. Second, epigenetic age clocks (Horvath, PhenoAge, GrimAge, DunedinPACE) have real population-level associations and poor test-retest reliability at the individual level, with technical replicate differences of one to three years on some clocks. A single clock readout is not a measurement of your aging rate. A slope across four or more longitudinal draws, processed on the same platform in the same batch where possible, is closer to something you can reason about.

The link is real but weaker than the marketing suggests. One-carbon metabolism supplies S-adenosylmethionine, the universal methyl donor, so folate and B12 status genuinely constrain the substrate pool for DNA methyltransferases, and both nutrient deficiency and genotype can shift global and gene-specific methylation.6 The broader framing of genome-diet interaction, where a common variant only matters against a particular nutritional background, is the right mental model for this pathway.7 Folate status has been associated with other molecular phenotypes as well, including telomere length in peripheral blood mononuclear cells, with a non-linear relationship to plasma folate.8 And methylation of the MTHFR promoter itself has been reported to interact with C677T genotype in disease association studies, which is a good illustration of why genotype and methylation are complementary rather than redundant measurements.9

What does not follow is a prediction. Knowing your MTHFR genotype does not let you predict your DNA methylation profile, your homocysteine, or your response to a supplement. The measurement that answers “is my one-carbon metabolism running short of substrate” is biochemical: plasma total homocysteine, serum or red cell folate, serum B12 with methylmalonic acid or holotranscobalamin as a functional marker. Order those, and if they are abnormal, that is a conversation with a clinician, not a supplement purchase. Our view on sequencing is straightforward: get whole-genome sequencing once for the genotype layer, add methylation profiling as a repeated measure if you intend to build a time series, and let the blood biomarkers do the work of telling you what is happening right now.

Questions people also ask

Which vitamins should I avoid if I have MTHFR? We cannot give you a supplement recommendation, and the genotype by itself does not support one. The common claim that C677T carriers should avoid folic acid in favor of methylfolate is not established by the clinical literature on these variants, and folic acid fortification remains the intervention with the strongest population evidence behind it.1 If you want an answer grounded in your own biology, measure homocysteine, folate, and B12 and take those numbers to a physician.

What are signs you may have MTHFR? Everyone has MTHFR, the gene. Carrying a common variant in it has no reliable signs, which is why it is found by sequencing and not by symptoms. Elevated plasma total homocysteine is the closest thing to a signal, and it depends far more on folate and B12 status than on genotype.1

What are the symptoms of MTHFR deficiency? Severe MTHFR deficiency, the rare inborn error, presents with homocystinuria and progressive neurological disease including seizures, and is confirmed by enzyme activity and metabolite testing under specialist care.3 It is a distinct condition from carrying C677T or A1298C, and a consumer panel reporting those variants is not testing for it.

Is a DNA methylation test the same as an epigenetic age test? An epigenetic age estimate is one model fitted on top of a methylation measurement, usually a few hundred CpGs from an EPIC array. The underlying data supports far more than the clock, and the clock has substantially more technical noise at the individual level than its single-number output implies.4

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. Anetta Undas, Krzysztof Chojnowski, Anna Klukowska, et al. Determination and interpretation of MTHFR gene mutations in gynecology and internal medicine. Polish Archives of Internal Medicine, 2019. https://doi.org/10.20452/pamw.15039 ↩ ↩2 ↩3

  2. John W. Steele, Sung-Eun Kim, Richard H. Finnell. One-carbon metabolism and folate transporter genes: Do they factor prominently in the genetic etiology of neural tube defects?. Biochimie, 2020. https://doi.org/10.1016/j.biochi.2020.02.005 ↩

  3. Kristin E. D’Aco, David Bearden, David Watkins, et al. Severe 5,10-Methylenetetrahydrofolate Reductase Deficiency and Two MTHFR Variants in an Adolescent With Progressive Myoclonic Epilepsy. Pediatric Neurology, 2014. https://doi.org/10.1016/j.pediatrneurol.2014.04.005 ↩ ↩2

  4. Ruth Pidsley, Elena Zotenko, Timothy J. Peters, et al. Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for whole-genome DNA methylation profiling. Genome Biology, 2016. https://doi.org/10.1186/s13059-016-1066-1 ↩ ↩2 ↩3

  5. The ENCODE Project Consortium. An integrated encyclopedia of DNA elements in the human genome. Nature, 2012. https://doi.org/10.1038/nature11247 ↩

  6. Bhongir Aparna Varma, Srilatha Bashetti, Rajagopalan Vijayaraghavan, et al. FOLIC ACID, VITAMIN B12, AND DNA METHYLATION: AN UPDATE. Asian Journal of Pharmaceutical and Clinical Research, 2018. https://doi.org/10.22159/ajpcr.2018.v11i1.21892 ↩

  7. Patrick J. Stover, Marie A. Caudill. Genetic and Epigenetic Contributions to Human Nutrition and Health: Managing Genome–Diet Interactions. Journal of the American Dietetic Association, 2008. https://doi.org/10.1016/j.jada.2008.06.430 ↩

  8. Ligi Paul, Marco Cattaneo, Armando D’Angelo, et al. Telomere Length in Peripheral Blood Mononuclear Cells Is Associated with Folate Status in Men ,. The Journal of Nutrition, 2009. https://doi.org/10.3945/jn.109.104984 ↩

  9. Ahmed M. Zain, Khalil A. El-Halfaway, Ahmed A. Abdel Megeed, et al. Methylomic Changes in MTHFR Promoter Region, along with the Heterozygous C677T Polymorphism, Contribute to the Risk of Thrombotic Stroke. Journal of Molecular Neuroscience, 2025. https://doi.org/10.1007/s12031-025-02364-1 ↩