Blood transcriptomics reveals a Parkinson's disease signature and heterogeneous prodromal molecular profiles in isolated RBD

Background: Parkinsons disease (PD) has a prolonged prodromal phase, but minimally invasive molecular biomarkers distinguishing manifest PD from prodromal synucleinopathy remain insufficiently characterized. Isolated REM sleep behavior disorder (iRBD) represents a high-risk prodromal condition and provides an opportunity to investigate early blood-based transcriptional alterations. Objective: To identify peripheral blood transcriptomic signatures distinguishing healthy controls (HC), individuals with iRBD, and patients with PD, and to explore whether longitudinal iRBD samples exhibit movement toward a PD-like transcriptional state. Methods: Peripheral blood RNA-seq data were analyzed using harmonized metadata, DESeq2 differential-expression analysis, internally validated machine-learning models, PD-like projection, and integrated biomarker-panel prioritization. Independent baseline samples were used for cross-sectional differential-expression and machine-learning analyses. iRBD follow-up and post-conversion observations were excluded from baseline model development and reserved for exploratory longitudinal analyses. Results: Baseline analyses included 71 independent samples: 20 HC, 31 iRBD, and 20 PD. An additional 19 iRBD follow-up observations, including three post-conversion observations, were available for exploratory analyses. Differential-expression analysis identified 170 FDR-significant genes in PD versus HC and 85 in PD versus iRBD, compared with one FDR-significant gene in iRBD versus HC. Internal machine-learning validation showed stronger discrimination of manifest PD, with a best ROC-AUC of 0.883 for HC versus PD and 0.889 for iRBD versus PD. Discrimination between HC and iRBD was weak, with a best ROC-AUC of 0.584. PD-like projection scores were lowest in HC, highest in PD, and heterogeneous among baseline iRBD samples. Follow-up iRBD samples showed an exploratory upward shift in the mean PD-like projection score. Integrated prioritization produced a 24-gene PD candidate panel and a 24-gene exploratory iRBD panel, with genes in each panel supported by machine-learning feature-stability evidence and differential expression analysis. Conclusions: Manifest PD was associated with a distinct peripheral blood transcriptional signature, whereas iRBD-associated alterations were substantially weaker and more heterogeneous. The prioritized panels represent candidates for independent technical and external validation and should not yet be interpreted as clinically validated diagnostic or prognostic tests.