Aleksandra Beric*, Sarp Sahin, Santiago Sanchez, Zining Yang, Ravindra Kumar, Isabel Alfradique-Dunham, Jessie Sanford, Daniel Western, Bridget Phillips, John P. Budde, Richard J. Perrin, Paul T. Kotzbauer, Joel S. Perlmutter, Scott A. Norris, Carlos Cruchaga and Laura Ibanez
Washington University in St. Louis
aberic [at] wustl.edu
Abstract
Identifying non-invasive Parkinson’s disease (PD) biomarkers is crucial for improving disease diagnosis and monitoring. We analyzed transcriptomic profiles from four major PD studies to identify circulating PD-associated transcripts that could serve as biomarkers.
Differential expression (DE) analyses were conducted in four independent PD whole-blood RNAseq datasets: PDBP (N=1,604), PPMI (N=1,943), BioFIND (N=213), and WashU-MDC (N=583). Results were consolidated through meta-analysis and biologically contextualized via pathway analyses and integration with brain transcriptomic, and plasma and CSF proteomic datasets. The available PD GWAS resources were utilized to verify whether the DE transcripts originated from known PD-risk loci. We examined the clinical relevance by correlating DE transcripts with UPDRS-III and MoCA scores. Finally, we leveraged our findings to develop predictive models to distinguish between PD and healthy controls.
We identified 296 DE transcripts, 28 of which were transcribed from known PD-associated loci. Our findings showed significant overlap with dysregulated brain transcripts (p=3.60×10-19). Nearly half of the identified transcripts were dysregulated before symptom onset. Over a third (105) of the DE transcripts correlated significantly with both UPDRS-III and MoCA scores. Three DE-transcript based predictive models distinguished PD from healthy controls with a ROC AUCs of 0.727-0.733. and were capable of detecting PD transcriptomic signatures before symptom onset. Two transcripts, DLD and CD55, emerged as promising early stage PD biomarkers that were differentially expressed in whole blood, brain and CSF and were included in all predictive models.
Keywords: transcriptomics, Parkinson's-disease, predictive-modeling, biomarkers