Omics Studies
High-throughput omics technologies have transformed ME/CFS research by enabling unbiased, hypothesis-free discovery of disease-associated molecular signatures.
1 Genomics
Genetic studies suggest a modest heritable component to ME/CFS susceptibility:
- Twin studies: Concordance rates are higher in monozygotic than dizygotic twins, suggesting genetic contribution, though shared environment cannot be fully excluded
- GWAS: No genome-wide significant loci have been consistently replicated across studies, likely reflecting polygenic architecture, phenotypic heterogeneity, and insufficient sample sizes. The UK ME/CFS Biobank and DecodeME (n\(>\) 20,000) represent the largest genetic studies to date and may have sufficient power to detect common variants of small effect
- HLA associations: Several studies report associations with specific HLA alleles (particularly HLA-DRB1), consistent with an autoimmune component in a subgroup. Associations with immune-regulatory genes (cytokine promoter polymorphisms, KIR genes) have also been reported but not consistently replicated
- Rare variants: Whole-exome sequencing studies are in early stages. Rare variants in mitochondrial DNA, ion channel genes, and complement pathway genes have been reported in individual families but require validation in larger cohorts
2 Transcriptomics
Gene expression profiling has identified reproducible transcriptomic signatures:
- Immune gene dysregulation: Multiple studies report altered expression of immune-related genes, including upregulation of interferon-stimulated genes, downregulation of NK cell-associated transcripts, and dysregulation of NF-\(\kappa\)B signaling pathways
- Metabolic gene changes: Downregulation of genes involved in oxidative phosphorylation and upregulation of glycolysis-related transcripts are consistent with the metabolic shift documented by functional studies
- Exercise-induced changes: Transcriptomic profiling before and after exercise challenge reveals gene expression changes in ME/CFS patients that differ from healthy controls, including altered stress response and immune activation patterns persisting 24–48 hours post-exercise
- Single-cell transcriptomics: Emerging single-cell RNA-seq studies are resolving cell-type-specific expression changes that are averaged out in bulk profiling, revealing that immune cell subsets carry distinct transcriptomic signatures in ME/CFS
Certainty: 0.30. (Preliminary, internally cross-validated, not externally validated, not published. Single-cohort. Not peer-reviewed. Shared as research-in-progress by Amatica Health at the Renegade Research Roundtable, 23 July 2026.)
Amatica Health has presented preliminary findings from a whole-blood RNA sequencing cohort of 244 participants (159 patients, 85 healthy controls) with deep phenotyping, including 63% of patients housebound or more severely affected. RNA was extracted from PAXgene tubes, depleted of ribosomal and globin RNA, and sequenced on a NovaSeq X (150 bp paired-end, ~60M reads per sample). An elastic net (GLMnet) classifier trained on 80% of the data and tested on held-out 20%, with ten-fold cross-validation and pre-registered acceptance criteria, achieved a held-out AUC of 0.893 (95% CI: 0.845–0.94, \(p = 0.01\) after 97 permutation tests). Subgroup performance: ME/CFS alone AUC 0.903, Long COVID alone AUC 0.858. Baseline models confirmed the RNA signal was not driven by demographics (age + sex AUC 0.48), cell composition shifts (AUC 0.60), or technical variables (AUC 0.63). At a balanced operating point the model achieved ~92% sensitivity and ~78% specificity. Performance degraded when specificity was tightened to 95% (sensitivity dropped to ~39%), so the classifier is not yet at clinical diagnostic thresholds.
Consequence: If externally validated at similar performance, this would represent the largest working RNA-based diagnostic classifier for ME/CFS — but the finding is preliminary, the cohort is single-site, and external validation is pending. No clinical application exists today.
Falsifiable prediction: External validation in an independent cohort with matched sequencing protocol should yield held-out AUC \(≥\) 0.80. If external validation AUC falls below 0.70, the internal AUC was driven by site-specific or batch-specific confounds rather than biology. The forthcoming paper and leave-one-batch-out testing (with 52 new samples added August 2026) will provide the first test.
Limitations: Internally cross-validated only — no external cohort validation yet. Single-site, single-protocol. Panel stability required tuning iterations. The blog post reports a paper in preparation but no manuscript or preprint was available as of July 2026. At 159 patients this is large for the field but small by clinical biomarker standards; performance at larger sample sizes is unknown.
Differential expression analysis (limma-voom) from the same cohort identified over 3,000 differentially expressed genes in the combined Long COVID + ME/CFS group, approximately 2,125 in ME/CFS alone, and approximately 1,472 in Long COVID alone. Pathway-level analysis (Reactome, fgsea) revealed two coordinated shifts at the cohort level. The B-cell and antibody arm appeared reduced, with lower activity in B-cell receptor genes, immunoglobulin genes, and master transcription factors defining B cells. Neutrophils showed a signature consistent with emergency granulopoiesis — raised activity in maturation, activation, and pathogen-response genes — alongside reduced neutrophil cell-type proportion estimates, making the activation signal paradoxical and potentially informative.
Certainty: 0.25. (Preliminary, confounded with sequencing batch for early pilot samples, cluster membership not stable. Not externally validated. Not published.)
Clustering patients on curated immune gene panels (B-cell, NK-cell, NLRP3/inflammasome, interferon/STAT, monocyte, T-cell, complement, mast-cell, TGF-\(\beta\), oxidative stress, RBC/heme) revealed a continuum of immune-inflammatory tone rather than discrete clusters. When a two-group split was forced, the groups separated into one immune-low group (~81 patients) and one immune-high group (~78 patients), but cluster membership was not stable under probing. Across compartments, one group sat roughly half a standard deviation below the cohort mean and the other roughly half a standard deviation above. The spread was visible in both Long COVID and ME/CFS, with samples from each end distributed across diagnostic labels. Two caveats apply: the split itself was unstable (the data is more consistent with a continuous spectrum than two distinct populations), and early pilot-batch samples are confounded with the immune-high group (a batch effect that has not yet been disentangled).
Consequence: If the immune continuum finding holds after batch-effect resolution, it challenges subgroup-based stratification approaches that assume discrete immunological clusters in ME/CFS. A continuum would imply that treatment-by-subtype logic needs continuous biomarkers rather than discrete groupings. The finding is preliminary and the batch confound must be resolved first.
Falsifiable prediction: After batch-effect correction and expansion of the pilot sample set, independent replication should fail to find a stable two-cluster split and instead confirm a unimodal or shallow-bimodal immune-state distribution. If a discrete clustering solution is robustly recovered after batch correction and external validation, the continuum interpretation is wrong and discrete immune subtypes exist.
Limitations: Batch confound (pilot samples cluster in immune-high). Unstable clustering. Single-cohort, internally only. No external validation. Curated gene panels may miss important pathways; exhaustive pathway discovery was not reported.
3 Proteomics
Proteomic studies complement transcriptomics by capturing post-translational regulation and secreted factors:
- Plasma proteomics: Studies have identified altered levels of complement components, coagulation factors, acute-phase proteins, and extracellular matrix remodeling enzymes in ME/CFS plasma
- CSF proteomics: The NIH study and others have profiled cerebrospinal fluid proteins, identifying alterations in neurotransmitter metabolism, neuroinflammatory markers, and blood–brain barrier integrity proteins
- Post-translational modifications: Oxidative protein modifications (carbonylation, nitrosylation) are elevated in ME/CFS, consistent with oxidative and nitrosative stress. Altered protein phosphorylation patterns in immune cells reflect dysregulated signaling cascades
4 Metabolomics
Metabolomics has produced some of the most striking and replicated findings in ME/CFS:
- Naviaux et al. (2016): Identified a hypometabolic signature in ME/CFS, with widespread reductions in amino acids, lipids, and nucleotides consistent with a dauer-like state of metabolic hibernation (Naviaux et al. 2016). The metabolic profile discriminated ME/CFS patients from controls with \(>\) 90% accuracy
- Tryptophan–kynurenine pathway: Multiple studies document a shift from serotonin synthesis toward kynurenine production, driven by inflammation-induced indoleamine 2,3-dioxygenase (IDO) activation. This has implications for mood, pain, and cognitive function (Chapter Neurological and Neurocognitive Dysfunction)
- Amino acid depletion: Reduced levels of branched-chain amino acids, tryptophan, and other essential amino acids are consistently reported, suggesting impaired anabolic metabolism
- Lipid abnormalities: Altered phospholipid profiles, reduced plasmalogens, and abnormal sphingolipid metabolism have been documented, consistent with membrane dysfunction and impaired mitochondrial lipid metabolism
5 Lipidomics
Dedicated lipidomic profiling has revealed ME/CFS-specific lipid signatures:
- Phospholipid remodeling: Altered phosphatidylcholine-to-phosphatidylethanolamine ratios suggest membrane instability. Reduced plasmalogens—antioxidant ether lipids concentrated in mitochondrial and cell membranes—may contribute to both oxidative vulnerability and mitochondrial dysfunction
- Eicosanoid imbalance: Altered prostaglandin and leukotriene profiles reflect immune dysregulation. Elevated prostaglandin E2 may contribute to pain sensitization and vasodilation
- Sphingolipid changes: Altered ceramide and sphingomyelin levels have been reported, with potential implications for immune cell signaling and apoptosis
6 Microbiomics
Gut microbiome studies have documented consistent alterations in ME/CFS (see also Chapter Gastrointestinal and Microbiome Dysfunction):
- 16S rRNA sequencing: Reduced microbial diversity and altered community composition, including reduced butyrate-producing bacteria (Faecalibacterium prausnitzii, Roseburia) and increased pro-inflammatory species. These changes are consistent with the gut barrier dysfunction and immune activation documented in ME/CFS
- Shotgun metagenomics: Functional profiling reveals altered metabolic capacity of the ME/CFS microbiome, including reduced short-chain fatty acid production and altered tryptophan metabolism—linking gut microbial changes to the systemic metabolic abnormalities detected by metabolomics
- Gut–brain axis: Microbial metabolites (short-chain fatty acids, tryptophan derivatives) directly influence CNS function through vagal signaling and systemic circulation, providing a mechanistic link between gut dysbiosis and neurological symptoms