A 34-circRNA blood biomarker panel outperformed plasma pTau217 in identifying Alzheimer's disease in this study.
The circRNA panel detected disease-related changes years before symptoms appeared, suggesting potential for earlier risk prediction.
People positive for both the circRNA panel and pTau217 had a higher risk of developing Alzheimer's than those negative for both.
Combining the circRNA panel with pTau217 improved diagnostic accuracy, suggesting the two biomarkers may complement each other.
The circRNA panel appeared relatively specific to Alzheimer's disease, with limited detection of Parkinson's, Lewy body, and frontotemporal dementia.
The findings were replicated in two independent cohorts but require larger studies before the test can be used in clinical practice.
Alzheimer's disease is usually confirmed through cerebrospinal fluid testing or costly PET imaging. A new Nature Medicine study reports that blood-based circular RNAs, a stable and often overlooked class of molecules, may emerge as an important addition to current Alzheimer's blood biomarkers after further validation.1
The results draw on more than 1,200 participants at Washington University's Knight Alzheimer Disease Research Center, replicated in two independent cohorts.1 If validated in future studies, such blood-based biomarkers could help identify people at risk earlier, enabling enrollment in clinical trials and potentially allowing disease-modifying therapies and other interventions to begin before significant brain damage occurs.
Circular RNAs are RNA molecules whose ends join into a closed loop rather than staying linear.1 This makes them unusually stable, resistant to degradation in blood, and able to cross the blood-brain barrier, with high expression in brain tissue.1 Unlike many RNA molecules that break down quickly in blood, circular RNAs remain stable for longer, making them attractive candidates for blood-based diagnostic tests.
Researchers screened over 1,600 circRNAs using two independent detection pipelines and identified 34 significantly linked to Alzheimer's status, including circDNAJC6, circMBOAT2, and circPICALM, a gene already tied to Alzheimer's risk.1 Most signals held up independently of their parent genes, suggesting the circular form itself carries disease-relevant information.1
The 34-circRNA model classified biomarker confirmed Alzheimer's status with an AUC of 0.945 (An AUC closer to 1.0 indicates better diagnostic accuracy.), compared to 0.877 for plasma pTau217.1 Combining both markers raised accuracy to 0.967.1 Performance held up across sexes and APOE4 carrier status, suggesting the result is not driven by a single subgroup.1 Against PET-confirmed amyloid rather than clinical status, accuracy was more modest at 0.757, improving to 0.931 with pTau217 added, suggesting the panel captures overall Alzheimer's biology better than amyloid buildup alone.1
| Biomarker | Sample Type | Invasiveness | Best At | What It Detects |
|---|---|---|---|---|
| CSF Aβ42/pTau181 | Cerebrospinal fluid | Invasive | Confirming amyloid/tau status | Amyloid-beta and tau pathology |
| Amyloid-PET | Brain imaging | Non-invasive, costly | Visualizing amyloid plaques directly | Amyloid plaques in the brain |
| Plasma pTau217 | Blood draw | Minimally invasive | Early amyloid pathology | Tau pathology associated with Alzheimer's disease |
| 34-CircRNA panel | Blood draw | Minimally invasive | Broader AD biology, progression risk | Multiple Alzheimer's-related biological changes reflected by circular RNAs |
Among cognitively unimpaired participants followed over time, the circRNA model predicted progression to symptomatic Alzheimer's better than pTau217, with a hazard ratio of 2.92 versus 1.81, meaning roughly triple the progression risk for high scorers, and a 5-year AUC of 0.870 versus 0.676.1 Combining both markers sharpened risk stratification: only 15 percent of double-negative participants progressed, versus 84 percent of double-positive participants.1
circRNA levels began shifting meaningfully 2 to 4 years before clinical onset, though this rests on small subgroups of just 5 to 33 participants per interval and needs confirmation in larger samples before the timing can be considered settled.1
Broader generalizability remains unconfirmed until tested in larger, more diverse cohorts.1
The panel showed low accuracy for Parkinson's disease, Lewy body dementia, and frontotemporal dementia versus controls, indicating relative specificity to Alzheimer's rather than neurodegeneration broadly.1 Three circRNAs did show weak associations across all four conditions, pointing to a small shared signal worth further study.1
Plasma pTau217 tracks amyloid pathology reliably, changing up to 15 to 20 years before symptom onset.3 But anti-amyloid therapies can lower pTau217 levels without a matching cognitive improvement, limiting its usefulness once treatment begins.2 This strengthens the case for biomarkers like circRNA that appear to reflect broader neurodegeneration rather than amyloid burden alone.1
These findings remain observational and cross-sectional. The study authors state that prospective validation in larger, more diverse populations is required before clinical use.1 Standardized assay development and regulatory review would typically follow.
Notably, the study's senior author and first author are co-inventors on a patent covering this biomarker, licensed to Circular Genomics; the senior author also serves on that company's scientific advisory board and holds equity, and two additional co-authors are Circular Genomics employees.1 This does not invalidate peer-reviewed, replicated findings, but is relevant context for readers evaluating how this technology might reach clinical use.
The 34-circRNA panel outperformed pTau217 in classifying Alzheimer's status and predicting progression.1 It appears Alzheimer's-specific and replicated across cohorts and, cautiously, across ancestries. It is not yet a clinical test, and several key findings rest on small subgroups. For now, it is a promising discovery-stage biomarker worth watching as larger studies unfold.
Phillips, Bridget, Jessie Sanford, Vaibhav A. Janve, et al. "Blood-Based Circular RNAs for Early Diagnosis of Alzheimer's Disease." Nature Medicine, published online July 1, 2026. https://doi.org/10.1038/s41591-026-04485-5.
Egan, Michael F., et al. "Randomized Trial of Verubecestat for Prodromal Alzheimer's Disease." New England Journal of Medicine 380 (2019): 1408–1420. https://www.nejm.org/doi/full/10.1056/NEJMoa1812840
Palmqvist, Sebastian, et al. "Discriminative Accuracy of Plasma Phospho-Tau217 for Alzheimer Disease vs Other Neurodegenerative Disorders." JAMA 324 (2020): 772–781. https://pubmed.ncbi.nlm.nih.gov/32722745/