The Epigenetic Clock as Diagnostic Tool
The consolidation variable \(overline(\mathcal{M})(t)\) in the ODE system represents the cumulative epigenetic inscription of the disease state. We propose that this variable is directly measurable using DNA methylation profiling of peripheral immune cells—operationalizing \(overline(\mathcal{M})\) as the deviation between disease-specific methylation patterns and healthy-baseline methylation at the same CpG sites.
ME/CFS methylation changes are bidirectional—both hypermethylation and hypomethylation at different loci—and the dominant direction may vary by tissue, cell type, and assay platform (Peppercorn et al. 2025) (Chalder et al. 2026) Wilfred C. de Vega, Vernon, and McGowan (2014). PTPRN2 hypomethylation correlates with cognitive symptoms; pericentromeric hypomethylation may derepress HSAT2 satellite repeats driving immune exhaustion; yet RRBS of PBMCs found 67.8% hypermethylated features in n=5 ME/CFS (Peppercorn et al. 2025). A scalar \(\mathcal{M}\) cannot represent this pattern faithfully.
Resolution: the vector model. The state variable \(\mathbf{\mathcal{M}} \in [0, 1]^n\) (defined in The Disease State ODE System) replaces the scalar with a per-locus vector. Consolidation depth is \(||\mathbf{\mathcal{M}} - \mathbf{\mathcal{M}}^\text{baseline}||\), capturing deviation in either direction. The gain and loss models are not competitors but complementary manifestations of a single underlying process — DNMT3B redistribution — which produces gain at some loci and loss at others. The vector model contains both as special cases (gain-dominant vs loss-dominant locus subsets) and unifies the paper’s competing formulations under one mathematical object.
Nucleosome mobility as the unifying molecular mechanism. The context-dependent functional effects of CpG methylation — activating transcription in gene bodies, repressing it at promoters — may all reduce to a single biophysical mechanism (Bonnet et al. 2026): methylation on ~200 bp DNA segments decreases nucleosome mobility. Reduced mobility at promoters prevents transcription factor access (repression); reduced mobility in gene bodies stabilises nucleosome positioning and facilitates transcription elongation (activation). This single mechanism explains why the same biochemical modification produces opposite functional outcomes depending on genomic location — consistent with a vector model where the sign of the effect \((m_i - m_i^\text{baseline})\) matters less than the per-locus context of the affected site. The therapeutic implication: global methylation-modifying drugs (DNMT inhibitors, methyl donors) affect nucleosome mobility at all methylated loci simultaneously, with location-dependent functional consequences that cannot be predicted from scalar methylation change alone.
Per-locus dynamics. Each locus \(i\) evolves according to:
\[ \frac{d m_i}{d t} = k_i^\text{meth} \cdot (1 - m_i) - k_i^\text{demeth} \cdot m_i - \beta_i \cdot (m_i - m_i^\text{baseline}) - \delta_i \cdot \Theta(m_i^\text{crit} - m_i) \]
where \(k_i^\text{meth}\) and \(k_i^\text{demeth}\) are locus-specific methylation and demethylation rates (functions of DNMT3B allocation at ProB vs ProA vs gene-region classes), \(\beta_i\) is the locus-specific tendency to return to baseline (passive maintenance), and the final term models irreversibility: \(\Theta\) is the Heaviside step function, \(\delta_i\) is a penalty term that activates when \(m_i\) falls below the critical density \(m_i^\text{crit}\), representing the additional barrier to de novo remethylation. For ProB repeats, \(k_i^\text{meth}\) is buffered by DNMT3B availability (constrained by the zero-sum redistribution across loci), so \(\sum_i k_i^\text{meth} ≤ K_\text{DNMT3B}^\text{total}\) — the locus-locus coupling constraint.
Histone mark coupling. At ProB repeats, CpG methylation loss is coupled to H3K9me3 loss; compensatory H3K27me3 gain (Polycomb) may partially rescue silencing but introduces a distinct chromatin state. The full state space is the tensor product \(\mathbf{\mathcal{M}} ⊗ \mathbf{\mathcal{H}}\) where \(\mathbf{\mathcal{H}} = (h_1, h_2, ..., h_n)\) tracks per-locus histone marks. \(B_\text{strength}\) depends on both methylation and histone state at ProB repeats. A full treatment exceeds current scope; the scalar histone index \(\mathcal{A}\) in Extended Subsystem Couplings provides a first approximation.
Tissue/cell-type heterogeneity. The vector model is tissue-indexed: \(\mathbf{\mathcal{M}}^\text{PBMC} ≠ \mathbf{\mathcal{M}}^\text{CNS} ≠ \mathbf{\mathcal{M}}^\text{muscle}\). Systemic interventions (methyl donors, SAMe, betaine) may differentially affect these tissue compartments. Methyl donors that penetrate the blood-brain barrier (methyl-folate, methyl-B12) may reach CNS loci; those that do not (some SAMe formulations) act only peripherally. Systemic remethylation may overshoot hypermethylated loci while correcting hypomethylated ones — a risk that per-locus targeting would avoid but current epigenetic drugs (global-acting HDAC inhibitors, DNMT inhibitors) cannot address. Tissue-specific models are desirable but data-constrained.
Clinical implication. The vector model replaces the binary “methyl-donors vs demethylators” therapeutic dichotomy with locus-specific targeting. The safest default strategy remains methyl-donor support (SAMe, methyl-folate, methyl-B12, betaine) — low-risk regardless of directionality — but the ultimate therapeutic goal is to restore \(\mathbf{\mathcal{M}} → \mathbf{\mathcal{M}}^\text{baseline}\) direction-correct at each locus: remethylation for hypomethylated ProB repeats (restoring heterochromatin) and demethylation for hypermethylated gene promoters (restoring expression). Current pharmacology cannot achieve this locus-specific precision, making the vector model both a more faithful representation of the biology and a call for the development of targeted epigenetic therapies.
Certainty. 0.55 for bidirectional pattern (documented across multiple studies). 0.40 for DNMT3B redistribution as the coupling mechanism (ProA/ProB framework (Bonnet et al. 2026) is a preprint, unvalidated for ME/CFS; the zero-sum constraint is mechanistically plausible but unmeasured). 0.45 for the irreversibility threshold \(m_i^\text{crit}\) (MeCP2/DNMT1 maintenance loop is well-established; Tiedemann 2024 provides direct experimental evidence that low-density CpG methylation is more vulnerable to maintenance disruption (Tiedemann et al. 2024); DNMT1 allosteric switch (Kimura and Sasaki 2012) further supports density-dependent maintenance. The quantitative threshold for any specific locus is unknown. 0.35→0.45: externally validated by Tiedemann 2024 NAR + Kimura 2012 review — general-population mechanistic evidence). 0.25 for the vector model providing better clinical guidance than the scalar model (direction-correct per-locus targeting is principled but pharmacologically unsupported). 0.15 for the tensor product \(\mathbf{\mathcal{M}} ⊗ \mathbf{\mathcal{H}}\) being operationally necessary (methylation-only models already capture the dominant signal; histone coupling may be a refinement rather than a structural requirement).
Provisionality. The ProA/ProB repeat framework (Bonnet, Hulo, Fourel et al. 2026 preprint) is a computational genomics hypothesis developed in cancer (Bonnet et al. 2026). Its extension to ME/CFS — including the DNMT3B redistribution mechanism, the \(B_\text{strength}\) derived variable, and the clinical implications — is our extrapolation and has not been endorsed by those authors. The Fourel paper is itself not peer-reviewed. The vector model is a mathematical container for this hypothesis; the hypothesis itself remains provisional.
The Horvath epigenetic clock (Horvath 2013) demonstrates that DNA methylation age can be precisely quantified from tissue samples. ME/CFS-specific methylation changes have been documented in multiple studies: de Vega et al. identified 1,192 differentially methylated CpG sites in ME/CFS PBMCs Wilfred C. de Vega, Vernon, and McGowan (2014), with subsequent work confirming glucocorticoid sensitivity-related methylation changes Wilson C. de Vega, Vernon, and McGowan (2021). Trivedi et al. identified additional methylation patterns with potential transposable element involvement (Trivedi et al. 2018). Helliwell et al. confirmed that methylation changes in ME/CFS reflect systemic rather than tissue-specific dysfunction (Helliwell et al. 2020). The GrimAge clock (Lu et al. 2019), which predicts mortality and morbidity from methylation data, could be adapted to predict ME/CFS disease trajectory specifically.
DNA methylation profiling of peripheral immune cells could serve as a direct measurement of \(overline(\mathcal{M})\)—quantifying epigenetic consolidation depth independently of symptom duration. Patients with genetically fast consolidation (DNMT3A/B gain-of-function polymorphisms) would have shorter intervention windows, identifiable before clinical deterioration signals the closing of that window.
Certainty: 0.40. Methylation arrays exist and ME/CFS methylation changes are documented across multiple studies Wilfred C. de Vega, Vernon, and McGowan (2014; Trivedi et al. 2018) Wilson C. de Vega, Vernon, and McGowan (2021), but no one has built a disease-duration-calibrated clock from these data. The concept is technically feasible with existing technology but has not been attempted.
Testable predictions:
- Methylation age of immune cells (deviation from chronological age) correlates with disease duration and severity, providing an objective biomarker independent of subjective symptom reports.
- Patients with DNMT3A gain-of-function variants show faster \(\mathcal{M}\) progression and earlier treatment resistance, identifiable through pharmacogenomic screening.
- A calibrated epigenetic clock predicts treatment response probability better than symptom duration alone—patients with low \(\mathcal{M}\) despite long duration (slow consolidators) should respond better than patients with high \(\mathcal{M}\) despite short duration (fast consolidators).
Limitations: Methylation changes in peripheral blood may not reflect tissue-resident immune cell or CNS epigenetic states. The causal direction—whether methylation changes drive disease persistence or merely record it—is not resolved by clock construction alone. Existing ME/CFS methylation studies used relatively small cohorts (\(n < 100\)); a disease-calibrated clock would require larger training datasets.