Belgian Researchers Identify Parkinson's Subtypes for Personalized Therapies at KU Leuven

Recent advancements in neuroscience at KU Leuven are reshaping our understanding of Parkinson's disease, a progressive neurological disorder affecting millions globally. Researchers at the VIB-KU Leuven Center for Brain & Disease Research have utilized innovative machine learning techniques on fruit fly models to pinpoint distinct molecular subtypes, opening doors to customized treatment strategies that could transform patient care.

This breakthrough highlights the power of interdisciplinary collaboration between biologists, computer scientists, and clinicians, underscoring Belgium's growing prominence in neurodegenerative disease research. By dissecting the disease's heterogeneity at a molecular level, the study challenges the traditional one-size-fits-all approach to therapy and promises more effective interventions tailored to individual biological profiles.

Understanding Parkinson's Disease: A Growing Global Challenge

Parkinson's disease (PD), the second most common neurodegenerative disorder after Alzheimer's, is characterized by motor symptoms such as tremors, muscle rigidity, bradykinesia or slowness of movement, and postural instability. Non-motor symptoms including sleep disturbances, cognitive impairment, and depression further complicate the condition. Dopamine-producing neurons in the substantia nigra region of the brain degenerate, leading to these manifestations.

Globally, approximately 12 million people lived with PD in 2021, with projections estimating over 25 million cases by 2050 due to aging populations and environmental factors. In Belgium, more than 65,000 individuals are affected as of 2026, with around 6,000 new diagnoses annually. The economic burden exceeds €1 billion yearly in direct medical costs and lost productivity, emphasizing the urgency for innovative solutions.

Current treatments like levodopa primarily manage symptoms but do not halt progression. Disease-modifying therapies remain elusive, partly because PD arises from diverse genetic mutations and environmental triggers, masking underlying molecular diversity.

The KU Leuven Breakthrough: Machine Learning Meets Fruit Fly Models

Led by Prof. Patrik Verstreken and first author Dr. Natalie Kaempf, scientists at the VIB-KU Leuven Center created 24 genetically precise fruit fly (Drosophila melanogaster) models mimicking human PD-associated gene mutations. These models recapitulate key PD features: age-dependent motor decline, loss of dopaminergic neuron innervation, and responsiveness to L-Dopa.

Rather than imposing preconceived categories, the team employed unbiased behavioral screening using high-throughput video tracking (Ethoscope system). They monitored locomotion, sleep patterns, and responses over days in thousands of flies. Machine learning algorithms, including non-negative matrix factorization (NMF) for pattern recognition and hierarchical clustering, revealed natural groupings without human bias.

This data-driven approach identified correlations between motor phenotypes and neuronal loss (R² = 0.53), validating the models' relevance to human disease.

Two Main Groups and Five Distinct Subtypes Emerge

The analysis clustered mutations into two primary functional groups, further subdivided into five subtypes:

• Group A (16 genes, proteostasis-focused): Involves disruptions in vesicle trafficking, retromer complex, lysosomes, and autophagy/proteostasis pathways maintaining cellular protein balance.
  • A1: Reduced sleep fraction (e.g., gba, iPLA2-VIA, DJ-1a/b).
  • A2: Reduced day sleep (e.g., Rab39, Vps35).
  • A3: Altered late-day sleep and bout increases (e.g., VAC14, Lrrk, Rme-8).
• Group B (8 genes, mitochondrial-focused): Pertains to energy production and mitochondrial health.
  • B1: Increased morning sleep, shortened latency (e.g., Pink1, park, Chchd2).
  • B2: Early morning sleep increases (e.g., nutcracker, Coq2).

Genetic interaction profiles via electroretinogram (ERG) assays confirmed intra-group synergies and inter-group antagonisms, bolstering the classifications.

Proof-of-Concept: Subtype-Specific Treatment Responses

A pivotal discovery was differential drug responses. Coenzyme Q10 (Q10), a mitochondrial supporter, rescued dopaminergic defects and motor function in Group B but not A. Conversely, R55, targeting trafficking pathways, benefited Group A, with some A3 overlap.

This explains past trial failures, like Q10's Phase III flop in mixed PD populations. Subgroup stratification could revive such compounds. Prof. Verstreken noted, "One drug to target different molecular dysfunctions in all Parkinson's doesn't exist."

The study, published March 10, 2026, in Nature Communications (read the full paper), provides a blueprint for precision medicine.

Building on Prior Subtyping Efforts

PD subtyping isn't new. Clinical schemas like tremor-dominant vs. postural instability/gait difficulty exist, but molecular insights lag. Recent 2024-2026 studies using imaging and biomarkers identified progression-based subtypes, yet few link genetics to function.

KU Leuven's work uniquely integrates familial mutations, prodromal symptoms like sleep changes, and therapeutic validation. It aligns with PPMI cohort efforts stratifying idiopathic PD, potentially via sleep/olfactory biomarkers.

Implications for Patients and Clinical Practice

For the 65,000+ Belgians with PD, this heralds biomarker-driven diagnostics—blood tests or imaging pinpointing subtypes for targeted therapies. Early intervention via non-motor signs could slow progression.

Clinical trials may shift to stratified designs, boosting success rates. In Europe, where PD prevalence nears 2 million, such advances could alleviate healthcare strains. For details on the VIB press release, visit VIB's announcement.

Boosting Neuroscience Research in Belgium

KU Leuven and VIB exemplify Belgium's neuroscience ecosystem, with Leuven Brain Institute fostering AI-biology synergies. The study involved KU Leuven's Computer Science and AI departments, highlighting higher education's role.

Funding from Aligning Science Across Parkinson’s (ASAP) underscores international collaboration. This positions Belgian universities as hubs for neurodegeneration research, attracting talent amid rising PD burden.

Career Opportunities in Parkinson's Research

This discovery fuels demand for neuroscientists, geneticists, and bioinformaticians. KU Leuven advertises postdocs in PD mechanisms, while VIB offers PhD programs in neuroscience. Skills in Drosophila genetics, machine learning, and high-throughput screening are prized.

Belgium's research ecosystem, with competitive salaries and EU grants, appeals to early-career academics. Explore roles in synaptic biology or mitochondrial dynamics for impactful contributions.

Future Directions and Challenges Ahead

Translating fly findings to humans requires validating subtypes in PPMI cohorts and GWAS risk genes. Develop subtype biomarkers and subtype-specific drugs. Challenges include idiopathic PD dominance and multi-pathway genes like GBA.

Optimism prevails: Prof. Verstreken envisions applying this to other polygenic diseases. By 2030, subtype-stratified trials could yield first disease-modifying therapies, extending quality life for millions.

For more on the study's reception, see Medical Xpress coverage.

Photo by Alexander Van Steenberge on Unsplash

Stakeholder Perspectives and Broader Impact

Patient advocates praise the shift to precision medicine, while pharma eyes repurposed drugs like Q10. Policymakers note economic savings from effective treatments. Academics celebrate unbiased ML's role in discovery.

In higher education, it inspires curricula integrating AI in biomedicine, preparing students for translational research.