Cross-cohort exosomal miRNA profiling identifies a robust Parkinson’s disease signature with multi-omics evidence for mitochondrial and immune pathology

Parkinson’s disease (PD) lacks reliable non-invasive biomarkers for early diagnosis. Peripheral blood-derived exosomes, which carry central nervous system (CNS)-derived cargo across the blood-brain barrier, represent a promising liquid biopsy platform. Here, we performed integrated miRNA and proteomic profiling of blood exosomes, leveraging multiple public datasets for biomarker discovery and an in-house cohort for independent validation.

Consensus feature selection integrating five methods identified robust biomarker panels evaluated by four classification models. A 69-miRNA consensus signature achieved strong PD classification performance, with Random Forest demonstrating the best generalization (AUC = 0.952), whereas the 63-protein proteomics signature showed limited cross-study reproducibility (best AUC = 0.611), attributable to inter-platform heterogeneity. Tissue origin profiling revealed that signature miRNAs derive from both CNS and peripheral immune compartments.

Functional enrichment across omics layers consistently highlighted mitochondrial energetics failure and suppression of complement and coagulation cascades. Deconvolution and independent single-cell RNA sequencing validation traced exosomal origins to a distinct immune-metabolic axis involving B cells and CD4 + T cells. Our study identifies a robust exosomal miRNA signature for PD classification and reveals the convergence of mitochondrial dysfunction and peripheral immune dysregulation as core pathological features, establishing a foundation for exosome-based liquid biopsy and mechanistic biomarker development.

The authors thank Professor Yamei Tang (Quanzhou Top-Talent Innovation Laboratory, Grant No. 00212006) for her support. Xianlong Wang was supported by Fujian Medical University (Grant No. XRCZX2017001) and the Natural Science Foundation of Fujian Province (Grant No. 2025J01843).

Chunnuan Chen was supported by the Natural Science Foundation of Fujian Province (Grant No. 2023J01104), the Scientific Foundation of Quanzhou City for High Level Talents (Grant No. 2023C005YR), the Joint Funds for the Innovation of Science and Technology, Fujian Province (Grant No. 2023Y9255), Fujian Health Science and Technology Plan Project (Grant No. 2024GGA039), and the Education and Teaching Research Project of Fujian Province for Undergraduate Universities (Grant No. Shufen Liu was supported by the Startup Fund General Project of Fujian Medical University (Grant No. 2025QH1140). These authors contributed equally: Shufen Liu, Sijie Wang, Xiaofang Ye.

Department of Neurology, Rare Disease Medical Center, The Second Affiliated Hospital of Fujian Medical University, Quanzhou, Fujian, China Shufen Liu, Xiaofang Ye, Meifeng Wang & Chunnuan Chen Department of Bioinformatics, School of Medical Technology and Engineering, Fujian Medical University, Fuzhou, Fujian, China Fujian Key Laboratory of Medical Bioinformatics, Fuzhou, Fujian, China Institute of Precision Medicine, Fujian Medical University, Fuzhou, Fujian, China Correspondence to Xianlong Wang or Chunnuan Chen. The authors declare no competing interests. Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.