Emerging Biomarker Candidates from Parallel Multi-Modal Studies (2025–2026)
Certainty: 0.35. (Novel ratio concept; requires paired sample validation; platelet hyperactivation in ME/CFS is provisionally supported.)
The serum/plasma MMP-9 discrepancy is normally dismissed as pure artifact. However, the magnitude of the differential is itself informative: in healthy individuals, serum MMP-9 is approximately 3–4× higher than paired plasma MMP-9 because platelets and leukocytes release their MMP-9 stores during clotting (Jung et al. 2008) (Olson et al. 2008). If ME/CFS patients have hyperactive platelets (a finding supported by some preliminary studies), their clotting-induced MMP-9 release should be larger, producing a higher serum/plasma MMP-9 ratio than controls.
This ratio — serum MMP-9 / plasma MMP-9, measured from the same venipuncture — could serve as a functional platelet activation readout in a research setting, but only if independently validated against established platelet activation markers (platelet factor 4, β-thromboglobulin, P-selectin). Unlike flow cytometry-based platelet activation tests, which require fresh samples and expensive equipment, the MMP-9 ratio requires paired serum and citrate plasma collection, standard ELISA, and division of two values. Citrate plasma collection requires precise tube filling and standardized centrifugation/time-to-freeze (less than 30 min) — research-accessible with protocol infrastructure, not deployable in routine primary care. The ratio has no meaning without independent confirmation of platelet hyperactivity; differential clotting speed between patients and controls (documented coagulation abnormalities in ME/CFS) could produce spurious ratio differences unrelated to platelet biology. This approach has precedent in chronic spontaneous urticaria, where elevated serum/plasma MMP-9 ratios correlate with platelet hyperactivity and mast cell activation.
Testable prediction. In a paired-sample design (n=50 ME/CFS, n=50 controls), the serum/plasma MMP-9 ratio will differentiate ME/CFS from controls (AUC >0.70), and correlate with established platelet activation markers (platelet factor 4, beta-thromboglobulin, P-selectin) and with mast cell activation markers (tryptase, histamine metabolites).
Limitations. Serum/plasma ratio has never been formally validated as a platelet activation assay. Sampling protocol (collection tube type, centrifugation speed and time, time-to-freeze) must be rigorously standardized. The ratio may be confounded by platelet count, medications affecting platelet function (aspirin, NSAIDs), and recent exercise or meals.
Multiple 2025–2026 studies have produced parallel evidence for biomarker candidates across immune, neurological, and vascular compartments. While none have yet achieved clinical validation, the methodological diversity of these studies is encouraging. These studies identify different biological signals in different cohorts using different platforms—this is complementary multi-modal evidence, not convergence on a single mechanism. True convergence would require independent groups identifying the same biomarker abnormality with different methods. SleepFM (Thapa et al. 2026, Nature Medicine, n=65,000) illustrates an alternative convergence pathway: a single model demonstrates that cross-modal physiological decoupling during sleep predicts disease onset across 130+ conditions with C-index \(>\) 0.80 (Thapa et al. 2026), suggesting that physiological coupling patterns (a multi-system signal) can serve as a robust transdiagnostic biomarker. This does not constitute independent methodological convergence — it is a single model applied to diverse conditions — but it demonstrates that a single physiological parameter (decoupling) can carry diagnostic information across diseases. This suggests that ME/CFS biomarker research should complement molecular approaches with physiological coupling measures. The historical failure rate of ME/CFS biomarkers (NK cytotoxicity, cytokine panels, various autoantibody signatures—all initially promising, none clinically validated) should temper expectations.
1 Neuroinflammation Imaging Biomarkers
Yu et al. (2026) demonstrated that the neuroinflammation imaging (NII) model applied to diffusion MRI can detect widespread white matter abnormalities in ME/CFS that conventional DTI misses NII-HR (cerebral edema), NII-RF (cellular infiltration), and NII-FF (axonal reorganization) metrics correlated with mental health, disability, and disease severity in 67 ME/CFS patients vs. 67 matched controls. The NII model requires only standard diffusion MRI sequences (no contrast agent, no special hardware), making it potentially translatable to clinical settings if validated in larger cohorts.
(Certainty: 0.40. Raw 0.40, population 1.00, discounted 0.40.) The Vienna experimental-hypoxia probe reported elevated resting thalamic lactate (Lac/tCr ~27% higher in ME/CFS than controls) alongside blunted metabolic reactivity to hypoxia (Bader et al. 2026). Elevated baseline brain lactate could therefore serve as a mechanism-based stratification biomarker for trials of tissue-oxygenation or mitochondrial interventions (e.g., hyperbaric oxygen therapy), identifying a biologically coherent “virtual-hypoxia” subgroup hypothesized to respond (Hadanny et al. 2024).
(Origin: brainstorm.) Replication status: Not independently replicated; single-site preprint. Severity applicability: Unknown — cohort not stratified by severity. Limitations: Thalamic Lac/tCr is a research MRS measurement, not a validated clinical assay; reproducibility and discrimination (AUC versus healthy controls) are untested. It is not a validated diagnostic — hypothesis-generating only.
Falsifiable prediction: Thalamic Lac/tCr will (a) separate ME/CFS from controls with AUC \(≥\) 0.70 in a replication cohort, (b) be reproducible across test–retest, and (c) predict differential response to an energy-targeted intervention in a stratified trial — refuted if it fails any of these.
Consequence: A brain-scan lactate measure could help identify a biologically defined “brain-energy deficient” ME/CFS subgroup and match those patients to energy-targeting treatments, providing an objective change endpoint instead of symptom questionnaires — but only after prospective validation.
2 Complement-Based Subgroup Stratification
Maya et al. (2026) identified a complement pQTL-defined inflammatory subgroup in ME/CFS with a high C3/low Bb profile, validated against UK Biobank fatigue phenotypes Complement protein levels (C3, C4, Bb, C3a, C5a) could serve as a stratification biomarker to identify patients who may benefit from complement-targeted interventions. The genetic basis of the finding (pQTLs) suggests this is a stable trait rather than a transient state.
3 Extracellular Vesicle miRNA Signatures
Seifert et al. (2026) identified hsa-let-7b-5p downregulation in EVs from post-COVID ME/CFS patients, correlating with fatigue, pain, and impaired physical functioning EV-based biomarkers are advantageous because they reflect inter-cellular communication and may capture compartment-specific pathology. However, the small sample size (n=12 vs 15) and lack of replication limit current utility.
4 Longitudinal Exercise Proteomics
Germain, Hanson, and colleagues (2025) demonstrated that the pattern of proteomic response to exercise—particularly persistent immune/metabolic/neuromuscular dysregulation during the recovery phase—distinguishes ME/CFS from sedentary controls The proteomic signature of PEM (suppressed T/B cell signaling, upregulated glycolysis, disrupted IL-17) captured at 24h post-exercise may serve as an objective PEM biomarker, which is critically needed for clinical trials.
5 NET Degradation and DNase Activity as Candidate Post-Viral Biomarkers
NET degradation capacity—the balance between NET production and DNase-mediated NET clearance—has emerged from COVID-19 research as a potential post-viral biomarker with direct relevance to ME/CFS. Garcia et al. (Garcia et al. 2024) demonstrated that the ratio of NET markers (MPO-DNA, H3cit, H3cit-DNA) to functional DNase activity was markedly elevated in severe and critical COVID-19 patients and correlated with CRP and neutrophil/lymphocyte ratio—standard clinical severity markers. Crucially, H3cit and H3cit-DNA levels predicted hospitalisation among ambulatory patients, suggesting NET/DNase imbalance is an early rather than late feature of severe disease (Garcia et al. 2024). Assay components. A clinically applicable NET degradation panel would measure: (1) NET remnants (MPO-DNA complexes, citrullinated histone H3) by ELISA; (2) functional DNase activity by residual dsDNA degradation assay; (3) NET/DNase ratio as an integrated imbalance metric; and optionally (4) DNase1 and DNase1L3 antigen levels to distinguish enzymatic deficiency from consumption. The Garcia functional DNase assay—measuring residual dsDNA after plasma incubation—provides a scalable, quantitative method suitable for large cohorts that avoids the technical demands of zymography or continuous fluorometric kinetic assays (Garcia et al. 2024). Complementary calprotectin (S100A8/S100A9) measurement offers a clinically accessible proxy for NET burden (\(r >= 0.745\) with H3-NET levels) (Hetland et al. 2022). ME/CFS applicability. No study has measured NET remnants, DNase activity, or NET/DNase ratios in ME/CFS patients. This is a notable gap given that: (a) ME/CFS shares thrombo-inflammatory features with long COVID; (b) herpesviruses implicated in ME/CFS triggering (EBV, CMV) are established NETosis inducers (Schönrich and Raftery 2016); (c) ME/CFS patients have documented neutrophil dysfunction that may include defective NET clearance; and (d) NET/DNase imbalance is genetically influenced via DNASE1 polymorphisms, providing a testable genetic susceptibility hypothesis (Chapter Genetic and Epigenetic Factors). The NET/DNase ratio could serve as a post-viral stratification biomarker distinguishing patients with ongoing thrombo-inflammatory NET pathology from those with predominantly metabolic or neurological phenotypes. Technical precedent. The Garcia et al. assay was applied to n=145 COVID-19 patients across three severity strata with robust statistical adjustment (age, sex, BMI) (Garcia et al. 2024). The key challenge for ME/CFS application is the lower expected NET burden in chronic post-viral compared to acute severe infection, requiring high assay sensitivity and potentially enrichment for patients with recent symptom exacerbation or known thrombo-inflammatory features.
6 Epigenetic and miRNA Biomarker Candidates from the PTPRN2/miR-153-3p Axis
A 2026 study by Chalder and Moreau introduces a multi-layer epigenetic biomarker framework for ME/CFS patient stratification Three measurement modalities span distinct regulatory levels: PTPRN2 saliva methylation: A hypomethylated CpG site in PTPRN2 (protein tyrosine phosphatase receptor type N2) survived multi-factor correction in 54 ME/CFS patients versus 21 sedentary controls and distinguished epigenetic subgroups. The saliva collection modality offers a critical practical advantage: samples can be self-collected at home by severely ill patients who cannot attend clinics Circulating miR-153-3p: Reduced blood miR-153-3p correlated with poorer delayed memory recognition in ME/CFS patients, providing a potential biomarker for the memory-specific cognitive symptom domain miR-153-3p has independent support as a neurologically relevant miRNA in hippocampal memory, amyloid-beta regulation, and neuroprotection across multiple model systems EpiSwitch 3D chromatin conformation: Hunter et al. (2025) applied a different epigenetic modality — 3D chromatin conformation — to ME/CFS patients (n=47) versus healthy controls (n=61), achieving 92%/98% sensitivity/specificity in the discovery cohort (Hunter et al. 2025). This convergent finding from an entirely different epigenetic measurement approach strengthens the case for biologically distinct patient subgroups. Two caveats apply: the cohort was severely affected (housebound) patients, and the reported specificity is against healthy controls only — the test has not been assessed against fibromyalgia, depression, or long-COVID fatigue, so it is not yet shown to be specific to ME/CFS as opposed to other fatiguing conditions.
None of the epigenetic biomarker candidates described here (PTPRN2 methylation, miR-153-3p, EpiSwitch panel) have been independently replicated in external ME/CFS cohorts. The Chalder 2026 primary finding derives from n=54 patients using a cross-sectional design with saliva proxy. The EpiSwitch result derives from a single retrospective vendor-affiliated study (Oxford BioDynamics) with no independent external validation. Reference ranges for clinical use do not exist. Pre-analytical variables (time of collection, saliva contamination, medication effects on methylation) are not established for these assays. These findings are research tools only — not diagnostic or prognostic biomarkers for clinical practice.
Consequence: None yet for patients — these are research-stage markers, not ready-to-use blood tests; their value today is in guiding further study, not clinical decisions.
7 Repetitive Element RNA Biomarkers: HSAT2 Detection in Biofluids
Satellite 2 (HSAT2) RNA represents an emerging class of biomarkers based on repetitive element detection in serum and plasma, with validated methods from cancer research immediately applicable to ME/CFS. Validated detection methods. Two robust methods for HSAT2 quantification in biofluids have been published:
TRAP-ddPCR: Tandem repeat amplification by nuclease protection combined with droplet digital PCR achieves high-sensitivity detection of serum HSATII RNA (Seimiya et al. 2023). This method uses nuclease protection to preserve repetitive sequences followed by ddPCR quantification, achieving high sensitivity and specificity in cancer discrimination (AUC ≥0.90)
Hybridization capture: Biotinylated probe hybridization capture for plasma cfDNA HSAT2 (Yörüker et al. 2026) enables size-selective detection of specific HSAT2 fragments (95 bp vs 114 bp), with fragment size distribution providing additional discrimination power Clinical validation status. HSAT2 detection has been validated in cancer cohorts (pancreatic cancer (Seimiya et al. 2023), colon cancer (Yörüker et al. 2026)) but not yet tested in ME/CFS or any fatigue-related condition. The methods are transferable: TRAP-ddPCR requires standard qPCR equipment plus droplet generation system; hybridization capture requires standard next-generation sequencing library prep. Cross-disease retrotransposon activation patterns. Retrotransposon activation provides a comparative framework across disease categories. In cancer, HSAT2 serves as a validated biomarker with established detection methods. In post-viral syndromes, endogenous retrovirus (HERV) activation has been documented during and after SARS-CoV-2 infection (Charvet et al. 2023), with HERV-W ENV protein expression correlating with disease severity (Charvet et al. 2023). However, direct HSAT2 measurements in Long COVID, fibrotic diseases, or other post-viral conditions represent a critical research gap (certainty: 0.45 — established for HERV, absent for HSAT2). The pattern suggests retrotransposon activation may be a pan-viral or pan-inflammatory response rather than disease-specific. If HSAT2 in ME/CFS mirrors the HSAT2 patterns seen in cancer (high levels correlating with disease stage), this would support a generalized retrotransposon activation mechanism. If HSAT2 patterns in ME/CFS differ (e.g., elevated but stable rather than progressive), this would suggest disease-specific regulatory dynamics. Exosome-mediated retrotransposon transfer. Evdokimova et al. (Evdokimova et al. 2019) demonstrated that HSAT2 and HERV-K RNAs are selectively packaged into extracellular vesicles and can be transferred to immune cells, inducing immunosuppression. Retroviral transcriptome dynamics in COVID-19 show similar stage-specific HERV activation patterns, suggesting that EV-mediated retrotransposon transfer may be a shared pathogenic mechanism across viral and fatigue syndromes. The ME/CFS-specific question is whether HSAT2-containing EVs drive the documented immune exhaustion (MDSC expansion, T-cell dysfunction) in the same way they do in cancer. Proposed ME/CFS biomarker validation study.
Phase 1 (Analytical validation): Establish reference ranges for serum HSAT2 RNA (TRAP-ddPCR) and plasma cfDNA HSAT2 (hybridization capture) in 100 healthy controls
Phase 2 (Case-control): Measure HSAT2 levels in 200 ME/CFS patients (100 post-viral, 100 non-post-viral) vs 100 matched controls; assess diagnostic accuracy (sensitivity, specificity, AUC)
Phase 3 (Subtype stratification): Correlate HSAT2 levels with (a) disease duration, (b) severity, (c) immune exhaustion biomarkers (MDSC, T-cell exhaustion), (d) viral serologies (EBV, HHV-6, SARS-CoV-2)
Phase 4 (Longitudinal): Track HSAT2 levels over 12 months in 50 patients to assess stability and correlation with symptom fluctuation Connection to pathogenic mechanisms. If HSAT2 is elevated in ME/CFS, it may serve as both biomarker and pathogenic mediator. Chapter ME/CFS Through the Lens of Universal Disease Mechanisms (Section Family 20: Inflammation Resolution and Lipid Mediators) describes the hypothesis that exosomal HSAT2 drives myeloid-derived suppressor cell expansion and T-cell exhaustion (Evdokimova et al. 2019). Chapter Energy Metabolism and Mitochondrial Function (Section Step 10: Mitochondrial Dynamics and Biogenesis) describes epigenetic HSAT2 activation pathways via HSF1, CTCF loss, and DNA hypomethylation. Biomarker validation studies should therefore include mechanistic readouts (immune phenotyping, exosome HSAT2 cargo) to distinguish whether HSAT2 is merely elevated or actively driving pathology.
Certainty: 0.30. (Internally cross-validated, not externally validated, not published. Single-cohort, 244 participants. Presented at Renegade Research Roundtable, July 2026.)
An elastic net classifier trained on whole-blood RNA-seq data (159 patients, 85 controls, 63% housebound) achieved held-out AUC 0.893 (95% CI 0.845–0.94, permutation \(p = 0.01\)), substantially outperforming cell-composition (AUC 0.60) and technical-variable (AUC 0.63) baselines. Pre-registered acceptance criteria were met after expression-floor thresholding and elastic-net tuning. At present the classifier reaches ~92% sensitivity/~78% specificity at a balanced operating point, well below the clinical threshold of ≥95% for both metrics. The cohort is the largest RNA-seq diagnostic attempt in ME/CFS to date but represents a single site with internal cross-validation only.
Consequence: Demonstrates proof of concept that whole-blood RNA-seq can separate ME/CFS from controls with a pre-registered, permutation-tested classifier. Moves the field from isolated gene-signature studies toward a validated analytical framework — but remains research-stage only until externally validated.
Falsifiable prediction: External validation of the classifier in an independent cohort will yield AUC \(≥\) 0.80. If external AUC falls below 0.70, the internally cross-validated result was driven by site-specific confounds. The leave-one-batch-out test with 52 additional samples (expected August 2026) is the first near-term test.
Limitations: No external validation. Not published or preprinted. Single-protocol. Panel stability required iteration. A paper is reportedly in preparation. The finding’s relevance at clinical biomarker thresholds (≥95% sensitivity + specificity) is unproven.
Certainty: 0.20. (Cluster membership unstable; confounded with batch for early pilot samples. Internally only. Not published.)
Curated-system immune gene clustering revealed a continuum of inflammatory tone across 159 patients rather than discrete immune subtypes. A forced two-group split produced immune-low (~81 patients) and immune-high (~78 patients) groups separated by roughly 1 standard deviation across compartments (monocytes, NLRP3, neutrophils, interferon, B-cells, T-cells, complement, mast cells, NK cells), but the split was not stable and is partially confounded with sequencing batch. The spread appeared in both Long COVID and ME/CFS, not specific to one diagnostic label.
Consequence: If confirmed, a continuum would suggest that biomarker-stratified trial designs assuming discrete immune subtypes may be premature — but the batch confound and instability mean this is a provisional signal only.
Falsifiable prediction: After batch correction, clustering on the full unbiased gene set (not curated panels) will fail to recover stable discrete clusters with silhouette scores above chance levels. If discrete clusters are robustly recovered, the continuum interpretation is wrong and immune subtypes exist.
Limitations: Unstable clustering. Batch confound for pilot samples. No external validation. Curated gene panels only. The data is more consistent with a continuous spectrum than discrete subgroups, but the batch effect has not been untangled.