Blood Multi-omics Reveal Persistent Molecular Disruption in PASC
Millions of people recover from acute SARS-CoV-2 infection only to remain symptomatic for months, and this study pushes the biology of post-acute sequelae of COVID-19 into sharper focus. In a cohort built around self-reported PASC, investigators profiled peripheral blood with mass spectrometry and next-generation leukocyte RNA sequencing, comparing 155 individuals with PASC against 100 acutely ill COVID-19 patients and 44 healthy controls. The central finding was not simply the presence of abnormal signals, but the persistence of a molecular state that failed to return to healthy baseline.
Persistent symptoms tracked with persistent biology.
The design matters because blood is one of the few clinically accessible tissues in long COVID research. Peripheral blood does not capture every organ-level process, but it offers a practical window into immune activation, lipid signaling, and chromatin regulation, all of which can influence fatigue, dyspnea, cognitive symptoms, and post-exertional worsening through systemic inflammation and altered cellular programming. By combining proteomic and transcriptomic readouts, the investigators were able to look beyond single biomarkers and instead examine the architecture of the post-infectious state.
Failure to Revert to Healthy Baseline Defines the Molecular Signal
The study linked PASC status to a broad failure to regain healthy molecular homeostasis. Rather than returning to the control pattern seen in uninfected subjects, the PASC group retained aberrations across multiple biological pathways, including sphingolipids and chromatin-associated processes. Those two domains are biologically plausible anchors for chronic illness: sphingolipids influence membrane structure, signaling, and inflammatory responses, while chromatin-linked changes can reflect longer-lived shifts in gene regulation after immune activation. In practical terms, those pathways suggest that long COVID is not a single biochemical defect but a layered disturbance affecting both signaling chemistry and transcriptional control.
The comparison groups strengthen the interpretation. Acute COVID-19 provides a reference for the inflammatory phase of infection, while healthy controls define the recovered baseline. The PASC group falling somewhere else entirely, and not simply between the two, supports the idea of a distinct post-acute state rather than lingering acute illness. That distinction has direct implications for biomarker development, because a useful diagnostic panel must separate chronic sequelae from the biology of recent infection.
Machine Learning Identified a Nine-Entity Diagnostic Signature
The most clinically concrete result came from the computational analysis. A machine-learning algorithm trained on participant data selected nine top entities that distinguished PASC with a mean precision-recall area under the curve of 0.97. This is a strong exploratory signal, but it should be read in the context of discovery-stage biomarker work rather than as a ready-made clinical test. Precision-recall performance is especially relevant when the positive class is difficult to define cleanly, as is often the case in syndromes with heterogeneous symptoms and variable timing after infection.
The biology behind the classifier matters as much as the numerical score. A small set of entities that can meaningfully separate PASC from acute COVID-19 and from health implies that the syndrome leaves a recognizable molecular fingerprint in blood. If validated, such a panel could help stratify patients who otherwise enter clinic with nonspecific complaints, overlapping fatigue syndromes, and inconsistent laboratory findings. For now, the model serves as a map of candidate signals rather than a bedside diagnostic.
The study also generated an interface tool to support rapid access to the dataset and encourage further analysis. In a field where many long COVID studies remain siloed, data accessibility is not a minor feature. Shared multi-omic resources allow other groups to test reproducibility, compare cohorts, and examine whether the same entities persist across institutions, disease severities, or symptom clusters.
Why Sphingolipids and Chromatin Changes Matter at the Bedside
The pathway findings point toward mechanisms that could explain chronic symptom burden. Sphingolipid dysregulation can affect immune signaling and membrane dynamics, potentially sustaining inflammatory crosstalk long after viral clearance. Chromatin-associated changes suggest that leukocytes may remain transcriptionally reprogrammed, which could help explain prolonged immune irregularity even when standard clinical tests appear unrevealing. Those mechanisms are not mutually exclusive; they may reinforce one another, with altered lipid signaling feeding into persistent gene-expression changes in circulating immune cells.
That is a clinically relevant frame for PASC because patients often present with symptoms that are real, disabling, and difficult to anchor to routine laboratory panels. A molecular signature in blood does not replace clinical evaluation, but it can give structure to a syndrome that has often been dismissed as too diffuse to study. It also creates a path for future work to determine whether distinct symptom subtypes map onto different multi-omic patterns.
Exploratory Data, Not Yet a Clinical Test
Several limits remain embedded in the study design itself. The PASC cohort was defined by self-reported status, which is practical for recruitment but introduces variability in symptom definition and timing. The abstract does not provide the exact symptom spectrum, illness severity distribution, or duration of follow-up, so the durability of the observed signatures across the full clinical course cannot be assessed from this report alone. The machine-learning result is also exploratory, and high apparent performance in a training context still requires external validation before any clinical use can be considered.
The study does, however, establish a credible framework for what long COVID biomarker research should look like next: multi-omic, comparative, and mechanistically grounded. It also provides a data resource large enough to support secondary analyses, which may be where the field starts to separate broad PASC biology from the more specific patterns that matter for prognosis, classification, and eventual therapeutic targeting. The unanswered question is not whether blood carries a signal. It does. The more important question is which part of that signal will hold up in independent cohorts and translate into clinically useful stratification.