Founder's role line (READ): "Establishes and removes DNA cytosine methylation, shaping the methylome and gene-expression programs." Key components (READ): DNMT1, DNMT3A/B, TET1/2/3
This is the strongest node in the slice and also the most contested. It is written so that it cannot be quoted as "fasting reverses aging."
DRY status
DRY: none located. Queries run:
"dry fasting" DNA methylation epigenetic OR immune OR VEGF OR Myc OR EZH2 human study(English)сухое голодание метилирование ДНК иммунитет лимфоциты исследование Филонов(Russian)Ramadan fasting DNA methylation epigenetic age human study(English, RAMADAN-targeted)
MEASURED: the two opened human dawn-to-dusk dry-fasting studies measured no DNA methylation and no epigenetic clock. One of them reports MECP2 protein up (log2 3.54, P=0.011); MECP2 is a methyl-CpG reader, not DNMT or TET, and is stated here so it is not counted as machinery. The RAMADAN-targeted query located a thesis record on epigenome-wide analysis of Ramadan fasting in pediatric MASLD (Hamad Bin Khalifa University). That record was located and NOT opened. Retrieval debt. Absolute DRY and the methylome: nothing located.
Disconfirming and null first
This section is longer than the supporting section. That is the honest shape of this literature.
1. The epigenome-wide result in the only long-term randomized CR trial is NULL. Ramaker ME, Corcoran DL, Apsley AT, Kobor MS, Kraus VB, Kraus WE, Lin DTS, Orenduff MC, Pieper CF, Waziry R, Huffman KM, Belsky DW. "Epigenome-wide Association Study Analysis of Calorie Restriction in Humans, CALERIE Trial Analysis." J Gerontol A Biol Sci Med Sci. 2022;77(12):2395-2401. doi:10.1093/gerona/glac168. PMID 35965483. MEASURED: zero CpG sites reached FDR < 0.05 at 12 months or at 24 months. Top site at 12 months cg07728874 (CD3D), p = 4.05e-6; top site at 24 months cg12040931 (a chromosome 1 lncRNA), p = 2.5e-6. Neither survived correction. Genomic inflation lambda = 0.97 at 12 months and 0.99 at 24 months. READ, authors' own words: "CR did not result in individual CpG-site DNAm changes that reached epigenome-wide significance." MEASURED: DNMT and TET genes are not mentioned anywhere in the article. Not one of the founder's six key components for this node was reported as changed by 24 months of caloric restriction.
2. Four of five epigenetic clocks in that same trial were NULL. One moved. Waziry R, Ryan CP, Corcoran DL, et al. "Effect of long-term caloric restriction on DNA methylation measures of biological aging in healthy adults from the CALERIE trial." Nature Aging. 2023 Feb 9;3(3):248-257. doi:10.1038/s43587-022-00357-y. PMID 37118425. PMCID PMC10148951. Correction: Nat Aging 2023 May 9;3(6):753. MEASURED, intention-to-treat, standardized d (95 percent CI):
| Clock | 12 months | 24 months | Verdict |
|---|---|---|---|
| PhenoAge (PC) | -0.03 (-0.19, 0.12), P>0.50 | 0.05 (-0.11, 0.20), P>0.50 | NULL |
| GrimAge (PC) | -0.04 (-0.16, 0.07), P>0.40 | 0.05 (-0.07, 0.17), P>0.40 | NULL |
| DunedinPACE | -0.29 (-0.45, -0.13), P<0.003 | -0.25 (-0.41, -0.09), P<0.003 | significant |
| Horvath (PC) | not in main text | not in main text | reported as not significant |
| Hannum (PC) | not in main text | not in main text | reported as not significant |
MEASURED from the trial's own preprint version (Waziry et al., medRxiv 2021.09.21.21263912): across the Horvath, Hannum and PhenoAge clocks, ITT Cohen's d ranged -0.02 to 0.15 with p>0.147, and most estimates pointed in the direction of FASTER aging in the CR arm. Treatment-of-treated d ranged -0.05 to 0.17, p>0.139, and higher CR dose was mostly associated with faster clock-age increase. MEASURED: the 20 percent-CR instrumental-variable estimates for PhenoAge and GrimAge were d = -0.13 to 0.01, P>0.15. MEASURED: the DunedinPACE effect corresponds to a 2 to 3 percent reduction in pace of aging.
3. The clocks are not measuring the same thing, and this trial is the demonstration. One clock moved and four did not, in one sample, on one set of arrays, at the same two timepoints. MEASURED, authors' own limitations: no gold-standard measure of biological aging exists; the DNAm measures are incomplete summaries with technical limitations; "The measures do not isolate system-specific aging processes"; effect sizes were small and the analysis was post hoc. DunedinPACE is a pace-of-change measure trained on longitudinal decline in organ-system biomarkers; the first-generation clocks are chronological-age predictors. A result on one is not a result on the other. Any sentence of the form "fasting lowered biological age" is unsupported by this trial for four of the five measures it reported.
4. This was NOT fasting, and the dose was not delivered. MEASURED: prescribed CR was 25 percent; achieved CR was 11.9 percent (s.e.m. 0.7) in Waziry 2023, and 11.7 +/- 0.7 percent by doubly labeled water in Ramaker 2022 (19.5 percent in months 0 to 6, falling to 9.1 percent in months 6 to 24). READ, authors' own words: "On average, trial participants did not achieve the prescribed dose of 25% CR." MEASURED: some control participants also reduced intake. CALERIE is labeled CR in this atlas, not fasting, per the brief. No fasting modality has an equivalent trial.
5. In an actual prolonged fast, no differentially methylated position was detectable, and the chronological-age clocks went UP. Luo S, Knufinke M, Kinnaer C, Chapatte L, Nusslé S, Nusslé SG, Grundler F, Wilhelmi de Toledo F, Horvath S, Mesnage R, von Meyenn F. "Effects of long-term fasting: longitudinal epigenetic responses in humans." 2026. doi:10.64898/2026.06.12.731903. Preprint; peer-review status not stated on the page, and the Data Availability section says data will be released "upon publication." Design MEASURED: 32 healthy adults (16 women, 16 men) from 33 included, one excluded for acute SARS-CoV-2; 12 consecutive days of medically supervised Buchinger-style fasting (about 600 kcal on a transition day, then about 200 to 250 kcal/day of juices, broths and honey, at least 2.5 L/day fluid), then 4 days of reintroduction and a one-month follow-up. Whole-blood Illumina EPIC array at baseline, day 10 and one month. MEASURED: no differentially methylated CpG sites between baseline and day 10. MEASURED: baseline to one month showed small-to-modest hyper- and hypomethylation with no significant Gene Ontology or KEGG pathway enrichment. MEASURED, 12 clocks: at day 10 the chronological-age clocks PCHorvath1, PCHorvath2 and PCHannum showed modest increases, while DunedinPACE and PCDNAmTL decreased. At one month PCHorvath2 stayed elevated, SystemsAge and DamAge rose with delay, AdaptAge decreased, and several others partly returned toward baseline. DNAm PAI-1 and the SystemsAge Heart score increased at one month. MEASURED: greater maximum weight loss (about 5.9 kg by day 12) was associated with lower baseline epigenetic age acceleration, nominally significant for some Horvath clocks and PhenoAge, borderline for DunedinPACE (p = 0.0504), not significant for others. That is a baseline association, so it describes who loses more weight, not an effect of the fast. This is the single most important row in the slice. The modality here is PF, it is the modality closest to what dryfasting.com's readers mean by fasting, and its methylome result at day 10 is nothing. The caveat that cuts the other way and is stated here too: the Buchinger protocol provides 200 to 250 kcal/day, so it is PF by convention in this literature and not a water-only fast.
6. The human intermittent-fasting literature has not measured this node. Işık T. "Epigenetically Mediated Health Effects of Intermittent Fasting" / "Aralıklı Açlık Diyetlerinin Epigenetik Aracılı Sağlık Etkileri." Istanbul Gelisim University Journal of Health Sciences (IGUSABDER). 2025;26:722-737. doi:10.38079/igusabder.1529934. Review article, Turkish journal. MEASURED from the opened review: DNA methylation is discussed only in animal studies; DNMTs appear only in a mouse offspring study (where IF reduced hepatic DNA methyltransferase levels); TET enzymes are not mentioned at all; epigenetic clocks are not mentioned at all; and the human studies it surveys measured miRNAs or gene expression, not methylation. MEASURED nulls it reports: Wegman et al., double-blind randomized, 12 healthy adults, 3 weeks of IF, found no significant changes in aging or metabolism genes or oxidative-stress markers, with only a 2.7 percent SIRT3 rise. Madkour et al. found SIRT1 and SIRT3 downregulated during Ramadan fasting, which the review states conflicts with the expected fatty-acid-oxidation gain.
7. Unresolved numeric discrepancy between the preprint and the published CALERIE clock paper. MEASURED from the published paper: DunedinPACE ITT at 24 months d = -0.25 (-0.41, -0.09), P<0.003. MEASURED from the preprint text as extracted: a 24-month ITT DunedinPACE estimate of +0.24, that is, the opposite sign, alongside a 12-month -0.29. Treatment-of-treated estimates in the preprint were -0.41 and -0.35. The preprint's Table 3, which holds the confidence intervals, could not be opened. This conflict is reported unresolved. It may be a preprint-to-publication revision or an extraction artifact in the preprint's narrative text. It is not resolved by anything located here, and the published value is not asserted as correcting the preprint. Retrieval debt recorded. Also unresolved in the same source pair: Ramaker 2022's abstract gives n = 197 with DNAm data while its Methods give 595 samples from 214 individuals (142 CR, 72 AL); and Waziry's 24-month AL count is 65 in Figure 1 and 68 in Table 2. The pages do not resolve either.
8. A second internal tension inside Ramaker 2022, reported rather than smoothed. READ: having stated that no individual CpG site reached epigenome-wide significance, the authors also state that secondary analyses of CpG sets taken from prior EWAS showed "clear evidence of DNAm changes in response to CR," and the abstract closes "DNAm changes may contribute to CR effects on aging." A site-level null and a set-level positive in one paper. Both are on the record. No winner picked.
Graded rows
| # | Claim | Modality | Species / system | Tier | Grade | Source | Tag |
|---|---|---|---|---|---|---|---|
| 37.1 | 24 months of prescribed 25 percent CR, 11.9 percent achieved, produced zero epigenome-wide significant CpG changes at 12 or 24 months in 197 to 214 healthy adults | CR | human RCT, whole blood | H1 | B+ | Ramaker 2022, J Gerontol A 77(12):2395-2401 | MEASURED |
| 37.2 | In the same trial DunedinPACE fell: d = -0.29 at 12 months and -0.25 at 24 months, P<0.003, about a 2 to 3 percent slower pace | CR | human RCT, n=220 randomized, 197 with DNAm | H1 | B+ | Waziry 2023, Nature Aging 3(3):248-257 | MEASURED |
| 37.3 | In the same trial PhenoAge and GrimAge were null at both timepoints (all 95 percent CIs span zero) | CR | human RCT | H1 | B+ | Waziry 2023 | MEASURED |
| 37.4 | In the same trial Horvath and Hannum were not significant, d -0.02 to 0.15, p>0.147, most estimates pointing toward faster aging in CR | CR | human RCT | H1 | B+ | Waziry et al., medRxiv 2021.09.21.21263912 | MEASURED |
| 37.5 | 12 days of prolonged fasting produced no differentially methylated CpG sites at day 10 in 32 adults | PF | human single-arm, whole blood | H3 | C | Luo 2026, doi:10.64898/2026.06.12.731903, preprint | MEASURED |
| 37.6 | In the same fast PCHorvath1, PCHorvath2 and PCHannum increased at day 10 while DunedinPACE and PCDNAmTL decreased | PF | human single-arm | H3 | C | Luo 2026, preprint | MEASURED |
| 37.7 | One month after that fast, SystemsAge and DamAge had risen, AdaptAge had fallen, and DNAm PAI-1 and SystemsAge Heart were up | PF | human single-arm | H3 | C | Luo 2026, preprint | MEASURED |
| 37.8 | IF reduced hepatic DNA methyltransferase levels in mouse offspring | IF-ADF (as described by the reviewing author) | mouse, liver | A1 | C | Işık 2025 review, primary study located via the review and NOT opened | READ, secondhand through a review; retrieval debt |
| 37.9 | 16:8 time-restricted feeding produced hypermethylation in a mouse vascular dementia model | TRE | mouse | A1 | C | Işık 2025 review, primary study NOT opened | READ, secondhand; retrieval debt |
| 37.10 | DNMT1, DNMT3A, DNMT3B, TET1, TET2 and TET3, the founder's six key components, have not been measured in any human fasting or CR intervention located in this sweep | all located modalities | human | n/a | n/a | absence across Ramaker 2022, Waziry 2023, Luo 2026, Işık 2025, Mindikoglu 2022, Mindikoglu 2025 | MEASURED as an absence of measurement; absence is not evidence of no effect |
Conflict left unresolved, named on both sides
- Site level versus clock level. Ramaker 2022 (site level, null) against Waziry 2023 (one clock of five, moved). Same trial, same participants. Unresolved.
- Set level versus site level inside one paper. Ramaker 2022's own site-level null against its own CpG-set-level positive. Unresolved.
- Pace clock versus chronological clocks. DunedinPACE down in CR (Waziry) and down at day 10 of a prolonged fast (Luo), while Horvath and Hannum are null in CR (Waziry) and up at day 10 (Luo). Unresolved, and the two sides are measuring different constructs.
- Preprint versus published 24-month DunedinPACE sign. Unresolved, see item 7.
Open questions
- No randomized trial of any true fasting modality (TRE, IF-ADF, IF-5:2, PF, FMD, RAMADAN, DRY) with epigenetic clocks as a prespecified endpoint was located. The one clock result is from CR.
- No human study located measured DNMT or TET protein or activity under any fasting modality.
- Whether the Luo day-10 clock rise is an acute artifact of the fasted state (shifted blood cell composition) or a real transient acceleration is unmeasured. The study reports whole-blood arrays.
← 36. SWI/SNF chromatin remodeling All 41 nodes 38. Epithelial-mesenchymal transition (EMT) →
dryfasting.com has been published since 2014. This evidence record was compiled on 22 November 2026. Every row names the fasting modality the study actually ran, the system it was measured in and the grade. Where no source carries a DOI or PMID the row says so rather than supplying one. Nothing is sold on this site.