Thermo and the Michael J. Fox Foundation Map 5,500 Samples to Hunt Parkinson’s Biomarkers

Thermo Fisher Scientific and the Michael J. Fox Foundation have completed a large-scale protein analysis of a landmark Parkinson’s study — and made the data public.

Announced August 13, 2026, Thermo analysed roughly 5,500 research samples from the Parkinson’s Progression Markers Initiative (PPMI), the foundation’s flagship study launched in 2010, using the Olink Explore HT platform.

The diagnostic problem

Parkinson’s is diagnosed largely by symptoms — a clinician observes tremor, rigidity, slowed movement and postural instability, and makes a judgement.

Two consequences follow. Diagnosis arrives after significant damage: by the time motor symptoms appear, a substantial proportion of the dopamine-producing neurons in the affected brain region have already been lost. Any treatment aiming to protect those neurons is therefore arriving late by construction.

And clinical diagnosis is imperfect. Several conditions produce similar features, and post-mortem studies have found meaningful misdiagnosis rates even among specialists.

The heterogeneity problem

Patients also vary widely, which makes one-size-fits-all drug development hard.

Some patients progress slowly over decades with predominantly tremor; others decline rapidly with early cognitive involvement. Some develop dementia; many do not. Response to treatment differs, and so does the pattern of non-motor symptoms — sleep disorder, constipation, loss of smell — that frequently precede movement problems by years.

Whether these represent one disease with variable expression or several distinct conditions sharing a clinical appearance is genuinely unresolved. If the latter, trials enrolling everyone with a Parkinson’s diagnosis would dilute any treatment effect confined to one subtype — which may be part of why so many neuroprotection trials have failed.

What proteomics offers

The Olink platform is a high-throughput method measuring thousands of proteins at once from a small sample.

Proteins are a useful layer to examine because they are what genes actually produce and what cells actually do. Genetic variants indicate predisposition; protein levels reflect current biological state, including the consequences of environment, disease process and time.

Measuring thousands simultaneously in thousands of samples generates the kind of dataset in which patterns distinguishing subtypes might emerge — patterns no single measurement would reveal.

Why PPMI is the right cohort

The value of the protein data depends entirely on the study it was drawn from, and PPMI has been running since 2010 collecting standardised clinical assessments, imaging and biological samples from patients and controls over time.

That longitudinal structure is what makes subtype identification possible. A protein pattern measured at one moment is a snapshot; the same pattern linked to how that patient progressed over subsequent years becomes a potential predictor.

Building such a cohort takes fifteen years and cannot be shortcut, which is why the analysis is being layered onto an existing study rather than a new one.

The open-data decision

Results are openly available in PPMI’s repository for researchers worldwide.

“This is what the Foundation’s global PPMI study was built to do,” said MJFF’s Samantha Hutten. Thermo’s Yan Zhang noted that “discovery is only the beginning in Parkinson’s research.”

Open release matters more than it might appear. A dataset of this size and quality analysed by one group yields whatever questions that group thinks to ask; released publicly, it is analysed by hundreds of groups with different hypotheses and methods.

It also allows replication, which proteomics badly needs — the field has produced many candidate biomarkers that failed to reproduce in independent datasets, and shared data is the most direct remedy.

The realistic expectation

The aim is to find biomarkers sorting patients into subtypes, tracking progression, and revealing the biological pathways behind the disease — groundwork for earlier diagnosis and precision-medicine treatments.

Worth being clear about the timeline. Identifying candidate biomarkers from a dataset is the first step; validating them in independent cohorts, demonstrating they predict something clinically useful, and developing them into usable tests takes years beyond that.

Proteomics has also disappointed before. High-throughput platforms generate enormous numbers of associations, most of which are chance findings or reflect confounders such as age, medication or inflammation rather than disease biology.

Why it is nonetheless the right work

The field is attempting to move from defining Parkinson’s by symptoms to defining it by biology — the shift oncology made decades ago, and Alzheimer’s research has been making more recently with amyloid and tau markers.

The prodromal window this could open

The most consequential application of biological markers in Parkinson’s is not better diagnosis but earlier diagnosis, and there is a well-characterised opportunity.

Parkinson’s has a long prodromal phase in which non-motor features appear years or decades before movement problems. Loss of smell, constipation, depression and particularly REM sleep behaviour disorder — physically acting out dreams — are recognised precursors, and a majority of people with that sleep disorder eventually develop a neurodegenerative condition.

Those features are non-specific individually, so they cannot support a diagnosis on their own. Combined with a biological marker, they might identify people in the prodromal phase with enough confidence to enrol them in prevention trials.

That is what the field needs most. Every neuroprotection trial to date has enrolled patients already diagnosed, meaning substantial neuron loss has occurred and there is less left to protect. Testing whether a treatment prevents Parkinson’s requires identifying people before they have it — which requires exactly the kind of biological definition this dataset is meant to enable.

That shift is a precondition for precision treatment. It is a resource rather than a therapy, and the kind of foundational work better therapies are built on. Research news.