Scientists have learned to recognize chronic fatigue by traces of cellular free RNA

A Cornell team has shown that a single vial of blood can provide a “molecular fingerprint” of myalgic encephalomyelitis/chronic fatigue syndrome (ME/CFS). They sequenced cell-free RNA (cfRNA) in plasma and trained machine learning models that distinguished patients from healthy (sedentary) individuals with ≈77% accuracy. The pattern suggested a malfunctioning immune system, a “loose” extracellular matrix, and signs of T-cell fatigue, with plasmacytoid dendritic cells (PCDCs) associated with the interferon response being particularly prominent. The work was published online August 11, 2025, in PNAS.

Background of the study

• The problem with no "tests." ME/CFS has no reliable lab test: the diagnosis is based on symptoms (post-exertional worsening, "brain fog," sleep disturbances, etc.) and the exclusion of other causes. Because of this, people go around in circles for years - there are few objective markers that a doctor could "hook on."
• It looks like a lot of things. ME/CFS complaints overlap with depression, anemia, thyroid dysfunction, autoimmune and post-infectious conditions, and in recent years, long COVID. There needs to be a biological fingerprint to help differentiate one from the other.
• Why did they try blood and cfRNA? Plasma contains fragments of RNA “dropped” by cells of different organs — cell-free RNA (cfRNA). It’s like a “black box” of the body: sets of such fragments can be used to judge which tissues and immune cells are activated, which pathways are “making noise” right now. This approach has already proven itself in other inflammatory and infectious conditions.
• What prevents us from seeing the signal? CfRNA is small, fragile, and ME/CFS patients are often sedentary — physical inactivity itself changes the molecular background. Therefore, it is important to build a rigorous laboratory pipeline (collection/storage/sequencing) and select the right control groups (including healthy but sedentary).

What was the aim of the work?

1. To understand whether ME/CFS has a persistent cfRNA signature in the blood.
2. Decompose the signal by sources: which cells/tissues contribute.
3. Identify biological pathways (immune dysregulation, extracellular matrix, signs of T-cell fatigue, etc.) that can be tested by other methods.
4. Building a machine learning model that can distinguish ME/CFS from controls is a step towards an objective test and future patient stratification.

Practical meaning

If the cfRNA signature is confirmed in large cohorts, it will produce:

• auxiliary diagnostic tool (not instead of the clinic, but to help);
• basis for ME/CFS subtypes (some are more “pro-interferon”, some are more pro-matrix/vessels, etc.);
• a path to targeted research and monitoring of response to interventions.

The idea is simple: instead of relying only on symptoms, read the body's systemic "event log" from the blood and extract from it a recognizable ME/CFS profile.

What did they do?

• They took blood from a group of people with ME/CFS and a matched group of healthy but sedentary participants (to avoid confusing the effects of the disease and inactivity). They isolated tiny fragments of RNA from the plasma that are released when cells are damaged and die—a sort of diary of what’s happening throughout the body. They then sequenced them and “taught” algorithms to find patterns of the disease. The result was >700 significantly different transcripts between cases and controls.
• Using the gene signature, the researchers "deconvoluted" the cfRNA and assessed which cells and tissues were sending the signal. They found differences in six cell types at once, with plasmacytoid dendritic cells, which produce type I interferons (a hint at a prolonged antiviral response), leading the way. Monocytes, platelets, and T-cell subtypes also changed.
• The cfRNA-based classifier achieved ≈77% accuracy—still low for a ready-made test, but a significant step forward toward objective diagnosis of ME/CFS.

Why is this important?

• There is currently no lab test for ME/CFS—the diagnosis is based on a combination of symptoms (severe fatigue, post-exertional worsening, “brain fog,” sleep disturbances, etc.), which are easily confused with other conditions. A blood “molecular cast” could give doctors a leg up—at least as an auxiliary tool at first.
• The approach is scalable: the same group of engineers has already used cfRNA to help differentiate Kawasaki disease, MIS-C, bacterial and viral infections in children—that is, it is a universal platform for complex diagnoses.
• For ME/CFS science, this is a step toward biomarkers of disease mechanics: the interferon axis, T-cell exhaustion, matrix disruption — all of which can be tested by other methods and integrated with proteomics/metabolomics. The field is already accumulating similar “puzzle pieces” (e.g., the role of oxidative stress and circulating microRNAs), and cfRNA adds a top-down view of the system.

Details that catch the eye

• >700 differential transcripts and focuses on pathways of immune dysregulation, extracellular matrix organization, and T-cell exhaustion are not just yes/no diagnostics, but hints at the biology of the process.
• The increase in signal from plasmacytoid dendritic cells (the main producers of IFN-I) is consistent with the hypothesis of a prolonged antiviral or "misguided" immune response in some patients.
• The team emphasizes that distinguishing ME/CFS from long COVID using cfRNA is potentially feasible and is a logical next step given the overlap between symptoms and mechanics.

Where is caution?

• This is not a ready-made analysis "from the clinic". 77% accuracy is a good start, but before the clinic, large, heterogeneous cohorts, external validation, comparison with other fatigue diseases and definition of pre-analytics standards (how to take/store blood) are needed.
• The control group is healthy sedentary people; it is important to check how the model works in real differential diagnoses in the office (depression, anemia, thyroid disease, autoimmune and post-infectious syndromes, etc.).
• cfRNA is a “summary” of the whole body; it is sensitive but also ambiguous. Therefore, interpretation must rely on independent data axes (proteomics, immunoprofiling, clinical).

What's next?

• Expand the dataset and refine the model to clinical metrics (AUC/sensitivity/specificity) in multicenter cohorts.
• To correlate cfRNA signals with symptom severity and post-exercise dynamics to approach patient stratification.
• Integrating cfRNA with already accumulated “omics” in ME/CFS and long COVID is the path to objective subtyping and targeted interventions.

Conclusion

Cell-free RNA has become the body's "black box": its patterns in the blood can be used to see the signature of ME/CFS, not just hear the symptoms. There will be no diagnostic test tomorrow, but the direction is clear: one test tube - a lot of biology, and doctors will have a chance to stop "feeling an elephant" blindly.