The Quest for a Universal Parkinson's Transcriptomics Signature is Derailed by Inherent Variability between Patients

Objective: To systematically evaluate the reproducibility and clinical utility of published blood mRNA-based gene signatures for classifying PD from healthy controls, and to uncover the pitfalls that limit their performances. Methods: We validated the classification performance of 13 gene signatures unique for PD (published 2015-2025) using the Parkinson's Progression Markers Initiative (PPMI) database. We further validated classification performance on data collected in a prospective clinical trial where demographic and clinical parameters were minimized and environmental confounders were strictly controlled. Sources of gene expression variability were studied by mean pairwise distance analysis. Results: Gene overlap between signatures was low (11/411, 2.7%) but statistically significant (P<0.001) and enriched with genes involved in lipid metabolism. The majority (10/13) of these signatures retained statistical significance when tested on the PPMI dataset, but their classification performance was modest (median AUC 59.7%), substantially lower than originally reported. Classification performance improved when comparing GBA1-associated PD to controls (median AUC 65.4%, P=0.006). A prospective trial (16 PD patients, 14 controls) with rigorous environmental standardization did not improve classification accuracy. Inter-individual expression variability is the dominant limiting factor in classifier performance. Variability was not impacted by long-term levodopa therapy. Interpretation: While blood transcriptomic signatures unique to PD are reproducible and may elucidate PD pathophysiology, their broad clinical utility remains severely limited by inherent inter-individual variability. Future research should prioritize biomarkers of divergence rather than the pursuit of a universal PD signature.