Summary: Researchers have developed an AI-driven blood test that can predict Parkinson’s disease up to seven years before symptoms appear. The test measures eight specific blood proteins and, in initial studies, diagnosed Parkinson’s with exceptionally high accuracy. This advance could enable much earlier intervention and clinical trials aimed at slowing or preventing the disease.
Parkinson’s disease is a progressive neurodegenerative disorder affecting nearly 10 million people worldwide. It is caused by the loss or dysfunction of dopamine-producing nerve cells in the substantia nigra, a brain region that controls movement. A pathological build-up of the protein alpha-synuclein is believed to contribute to this neuronal damage.
Key facts
- The test uses machine learning to analyse a panel of eight blood-based protein biomarkers associated with Parkinson’s disease.
- In the study, the model identified Parkinson’s patients with perfect accuracy and detected at-risk individuals up to seven years before motor symptoms began.
- Earlier diagnosis would make it possible to trial neuroprotective therapies sooner, aiming to slow or prevent the progression of Parkinson’s.
Source: UCL
A multidisciplinary team led by researchers at UCL and University Medical Center Göttingen developed a straightforward blood-based assay combined with artificial intelligence to predict Parkinson’s disease in its pre-motor phase. Their findings, published in Nature Communications, demonstrate that a targeted plasma proteomics panel, interpreted by machine learning, can distinguish people with Parkinson’s and identify many who will later develop the disease.
Today, most people with Parkinson’s receive dopamine replacement therapies only after clear motor symptoms appear, such as tremor, slowness of movement, altered gait and cognitive changes. Because neurons cannot be regrown, researchers emphasise the importance of identifying individuals at risk earlier, when therapies aimed at protecting surviving dopamine-producing cells would be most effective.
Professor Kevin Mills (UCL Great Ormond Street Institute of Child Health), senior author on the study, explained that as new disease-modifying treatments become available, diagnosing people before they develop symptoms is essential. He noted the aim to translate the test into routine large laboratory use within the health service, given adequate funding.
The study validated a multiplexed mass spectrometry assay and used a machine-learning algorithm to analyse eight proteins whose blood concentrations are altered in Parkinson’s. When tested on newly diagnosed motor Parkinson’s patients and matched controls, the algorithm accurately identified all Parkinson’s cases in the study cohort.
To assess predictive performance, the researchers analysed blood samples from 72 people with isolated REM sleep behaviour disorder (iRBD), a condition in which patients act out vivid or violent dreams. Longitudinal studies show that around 75–80% of people with iRBD will develop a synucleinopathy such as Parkinson’s. The machine-learning model classified 79% of the iRBD participants as having a biomarker profile matching Parkinson’s.
Over up to ten years of follow-up, the AI predictions tracked clinical conversions: the team correctly identified 16 individuals who later developed Parkinson’s, with predictions made as much as seven years before motor symptom onset. The researchers are continuing longitudinal follow-up to further confirm the test’s predictive accuracy.
Co-first author Dr Michael Bartl (University Medical Center Göttingen) and collaborator Dr Jenny Hällqvist (UCL Queen Square Institute of Neurology) emphasised that measuring eight blood proteins may allow identification of potential Parkinson’s patients several years in advance. Early identification would permit earlier administration of candidate therapies that might slow disease progression or prevent motor Parkinson’s altogether.
The protein markers identified reflect biological processes such as inflammation and impaired protein degradation, making them not only diagnostic markers but also potential therapeutic targets. Co-author Professor Kailash Bhatia and his team are evaluating the test in groups at higher genetic risk for Parkinson’s, including people with pathogenic variants in genes such as LRRK2 and GBA.
To broaden accessibility, the research team is seeking funding to develop a simpler finger-prick blood spot version of the test that could be mailed to a central lab. Such a format could support large-scale screening to identify at-risk individuals even earlier than the seven-year window reported in this study.
The research received support from an EU Horizon 2020 grant, Parkinson’s UK, the NIHR Great Ormond Street Biomedical Research Centre, and the Szeben-Peto Foundation. Professor David Dexter, Director of Research at Parkinson’s UK, described the work as an important step toward a patient-friendly blood test that could reduce reliance on more invasive sampling methods in research.
About this Parkinson’s disease research news
Author: Poppy Tombs
Source: UCL
Contact: Poppy Tombs – UCL
Image: The image is credited to Neuroscience News
Original Research: Open access. “Plasma proteomics identify biomarkers predicting Parkinson’s disease up to 7 years before symptom onset” by Kevin Mills et al. Nature Communications
Abstract
Plasma proteomics identify biomarkers predicting Parkinson’s disease up to 7 years before symptom onset
Parkinson’s disease progresses from a pre-motor phase—often marked by non-motor symptoms such as REM sleep behaviour disorder—to disabling motor symptoms. Objective blood-based biomarkers for these early stages are needed to enable timely intervention and to enrol at-risk participants into clinical trials aimed at slowing or preventing neurodegeneration.
The study validated a targeted, multiplexed mass spectrometry assay on plasma from recently diagnosed motor Parkinson’s patients (n = 99), two independent longitudinal cohorts of individuals with isolated REM sleep behaviour disorder (n = 18 and n = 54), and healthy controls (n = 36). A machine-learning model analysing eight proteins—granulin precursor, mannan-binding lectin serine peptidase 2, endoplasmic reticulum chaperone BiP, prostaglandin-H2 D-isomerase, intercellular adhesion molecule 1, complement C3, dickkopf WNT signalling pathway inhibitor 3, and plasma protease C1 inhibitor—accurately identified all Parkinson’s cases and classified 79% of pre-motor individuals up to seven years before motor onset. Several biomarkers correlated with symptom severity and point to molecular processes active in early disease stages. This panel could help select participants for trials testing therapies designed to slow or prevent motor Parkinson’s disease.