Background: Parkinsons disease lacks reliable early diagnostics and disease-modifying treatments. Blood-based biomarkers can facilitate early detection, symptom monitoring, and targeted therapy. Methods: We integrated gene expression and plasma proteomics data from two Parkinsons disease cohorts. Machine learning models were trained to classify disease status using protein, gene expression, and combined datasets. A severity score was derived by regressing protein levels against clinical motor ratings and tested longitudinally. Enrichment and network analyses identified biological context, and drug perturbation databases were queried for candidate therapies. Results: Proteomic models outperformed gene-based approaches and generalized well to external data. The severity score correlated with clinical burden and predicted future progression. Enriched pathways involved extracellular signaling, immune response, and post-translational regulation. Several compounds were identified as potential therapeutic candidates based on network targeting and reversal potential. Conclusions: Peripheral proteomic signatures offer classification, progression, and therapeutic insights in Parkinsons disease, supporting their biological relevance.
Nicholas C Minster (Thu,) studied this question.