Related Experiment Video
Updated: Jun 14, 2026
A High Throughput, Multiplexed and Targeted Proteomic CSF Assay to Quantify Neurodegenerative Biomarkers and Apolipoprotein E Isoforms Status
Published on: October 20, 2016
High-Throughput CSF Proteomics and Machine Learning to Identify Proteomic Signatures for Parkinson Disease
Kazuto Tsukita1, Haruhi Sakamaki-Tsukita2, Sergio Kaiser2
1From the Department of Neurology (K.T., H.S.-T., R.T.), Graduate School of Medicine, Kyoto University; Advanced Comprehensive Research Organization (K.T.), Teikyo University, Itabashi; Division of Sleep Medicine (K.T.), Kansai Electric Power Medical Research Institute, Osaka, Japan; Translational Medicine Department (S.K., P.S.-F.), Novartis Institutes for Biomedical Research, Basel, Switzerland; and Cardiovascular and Metabolism Department (L.Z.), and Neuroscience Department (M.M.), Novartis Institutes for Biomedical Research, Cambridge, MA. kazusan@kuhp.kyoto-u.ac.jp.
Researchers identified Parkinson disease (PD) specific cerebrospinal fluid (CSF) protein signatures using machine learning. This PD proteomic score (PD-ProS) aids in diagnosing PD and predicting disease progression in patients.
Area of Science:
- Neuroscience
- Proteomics
- Biomarker Discovery
Background:
- Parkinson disease (PD) diagnosis and progression monitoring require reliable biomarkers.
- Cerebrospinal fluid (CSF) proteomic analysis offers potential for identifying such biomarkers.
Purpose of the Study:
- To identify CSF proteomic signatures characteristic of Parkinson disease (PD).
- To evaluate the clinical utility of these signatures in diagnosing PD and predicting disease progression.
Main Methods:
- Observational study using data from the Parkinson's Progression Markers Initiative (PPMI) and LRRK2 Cohort Consortium (LCC).
- Aptamer-based CSF proteomic data quantifying 4,071 proteins were analyzed.
- Differentially expressed protein (DEP) analysis and least absolute shrinkage and selection operator (LASSO) were employed to derive a PD proteomic score (PD-ProS).
- PD-ProS was validated internally and externally, and its association with clinical progression was examined.
Main Results:
- A PD-ProS derived from 14 DEPs demonstrated good diagnostic performance (AUC 0.83) in differentiating non-genetic PD from healthy controls (HCs).
- Validation in an external cohort showed sustained diagnostic capability (AUC 0.75) even with a subset of proteins.
- PD-ProS effectively distinguished genetic PD from genetic prodromals and independently predicted cognitive and motor decline in PD patients.
Conclusions:
- High-throughput proteomics combined with machine learning can identify CSF proteomic signatures for PD.
- The developed PD-ProS shows potential as a diagnostic and prognostic tool for Parkinson disease, regardless of genetic status.
