Candidate biomarkers may help track Parkinson’s across disease stages

Researchers identified candidate biomarkers in spinal fluid and blood that showed different patterns among healthy controls, people in a prodromal group at risk of Parkinson’s, and those diagnosed with Parkinson’s disease, suggesting they may help track the disease across stages. The analysis included more than 1,100 participants.

“Our data-driven approach maps candidate biomarker trajectories across the [Parkinson’s] life cycle,” researchers wrote.

The study, “Proteo-metabolomic integration identifies stage-specific candidate biomarkers for Parkinson’s disease,” was published in npj Parkinson’s Disease.

Recommended Reading

Sept. 9, 2026 News by Andrea Lobo

New smartphone monitoring tool can help track Parkinson’s tremor

Early Parkinson’s changes can begin years before diagnosis

In Parkinson’s, the progressive loss of nerve cells that produce dopamine, a neurotransmitter that helps control movement, leads to motor symptoms such as slow movements, tremors, and rigidity.

One of the defining features of Parkinson’s is a long prodromal phase, when biological changes occur before a clinical diagnosis, sometimes lasting up to 20 years. However, by the time motor symptoms become clinically apparent, more than half of dopamine-producing neurons involved in movement are typically lost.

This highlights the need for biomarkers to detect and track the earliest biological signs of the disease.

Emerging evidence supports the use of molecular signatures that combine information from different biological fluids as biomarkers. In particular, protein and metabolite levels in blood and cerebrospinal fluid (CSF), which surrounds the brain and spinal cord, have shown biomarker potential, but their usefulness in predictive models is not yet well understood.

To address this, a team of researchers in Egypt analyzed proteomic data, which measure proteins, and metabolomic data, which measure small molecules, collected from CSF and blood samples through the Parkinson’s Progression Markers Initiative, an international observational study launched by The Michael J. Fox Foundation to identify biological markers of Parkinson’s progression.

The proteomic dataset included samples from 383 people with Parkinson’s, 101 in a prodromal group considered at risk of Parkinson’s based on clinical features, genetic variants, or other biomarkers, and 177 healthy individuals as controls. The metabolomic dataset included 604 Parkinson’s patients, 300 in the prodromal group, and 232 controls.

All biomarkers were measured at the first visit, and some participants had repeat measurements at follow-up visits extending up to 16 months for the progression analysis.

Initial comparisons found differences in 124 CSF metabolites, including 25 between the Parkinson’s and prodromal groups, and in 58 blood metabolites, including 36 between the two groups. Among proteins, the results showed 41 significant differences in CSF between controls and Parkinson’s patients, while no blood proteins met the study’s threshold for a significant difference in any comparison.

Protein and metabolite patterns differed across disease stages

In the metabolomic data, functional analysis highlighted pathways involving amino acids, the building blocks of proteins, and lipid (fat) metabolism. In the proteomic data, changes were mainly related to protein-binding functions and enzyme activity.

A statistical technique narrowed an initial pool of 1,138 protein and metabolite features to 318, creating a combined CSF-blood signature. Researchers then used those 318 features to train four machine learning models. Machine learning is a type of artificial intelligence that uses computer algorithms to learn from data and make predictions.

From this analysis, the researchers identified 21 candidate biomarkers that were selected by at least two machine learning models. These included established Parkinson’s biomarkers, as well as proteins and metabolites linked to key processes implicated in Parkinson’s biology, such as neuroinflammation, immune response, and neurotransmitter regulation.

To examine how the candidate biomarkers changed over time, the team tracked them across repeat visits. They narrowed the panel down to eight potential stage-specific biomarkers grouped into three exploratory subpanels.

The first group included four spinal fluid markers — the metabolite cadaverine alongside the proteins secretogranin, neuroendocrine protein 7B2, and LSAMP — that were considered potential early diagnostic markers because their levels differed significantly between the prodromal and control groups but showed no significant progression effect over time.

The second group included spinal fluid protein VGF and blood protein lumican as potential markers of the transition from the prodromal phase to clinical Parkinson’s.

The third group included blood proteins kininogen-1 and galectin-3-binding protein as potential progression markers. They showed some of the largest declines over time in the Parkinson’s and prodromal groups compared with controls.

Among the study’s limitations, the authors noted that the 16-month follow-up window is relatively short for a chronic, progressive disease. The findings also were not validated in a fully independent group of participants, and the researchers said the reported performance of the models may still be optimistic.

“This study illustrates that integrating [different types of molecular data] across multiple biofluids improves predictive performance for some [machine learning] frameworks,” the researchers wrote. “Ultimately, these findings underscore the value of a multi-analyte panel that strategically aligns biomarker selection with the biological compartment most reflective of central disease pathology.”