Plasma cell‐free RNA: a new non‐invasive biomarker for Parkinson’s disease

Abstract

AbstractBackgroundParkinson’s disease (PD) is the second‐most common neurodegenerative disorder, after Alzheimer disease (AD). We aimed to create a model that can predict PD with high accuracy and specificity using plasma cell‐free RNA (cfRNA) data.MethodWe extracted RNA from 172 plasma samples (92 PD cases and 80 controls; 48% female, and median age of 73) received from the Hope Center for Neurological Disorders at Washington University in Saint Louis. After library preparation, sequencing, processing, and stringent quality control data was split into training (60%) and testing (40%) populations, both of which included 53% PD cases, 48% female, and had median age of 73. We then performed differential expression analyses using DESeq2, adjusting the by age, sex, medication and other technical variables. Following data driven identification of informative transcripts using Lasso regression, we used Ridge regression to model disease status.ResultWe identified and replicated 639 significantly dysregulated genes. Through pathway analyses we found that dysregulated genes were enriched in PD terms, as well as other neurodegenerative and in known mitochondria pathways. After predictive model training, the best model for PD in the testing population included 14 genes and had an Area Under the Receiving Operating Curve (AUC) of 0.963 (0.930, 0.996). The model had low predictive power, with AUC of 0.553 (0.421, 0.686), in differentiating between controls and AD, FTD or DLB, suggesting that it is specific to PD. More evidence of its specificity was the good performance in distinguishing between PD and AD, FTD or DLB with AUC of 0.753 (0.674, 0.833) specificity.ConclusionPlasma cfRNA is a powerful tool for PD prediction, being a minimally invasive biomarker that could detect PD before onset of physical symptoms and without relying on imaging data. We are working on replicating our results in an independent dataset.