A groundbreaking study led by researchers from VIB and KU Leuven has revealed that Parkinson's disease comprises two primary categories and five distinct subgroups. Published in Nature Communications, this research utilized machine learning to uncover molecular differences that explain why a single treatment fails to benefit all patients. The findings challenge the traditional view of Parkinson's as a uniform disorder and pave the way for personalized therapies. Parkinson's disease currently affects millions globally and is diagnosed based on clinical symptoms such as movement difficulties and neurological decline. However, the condition can stem from mutations in numerous different genes, creating diverse underlying biological mechanisms. This genetic complexity has hindered drug development, as therapies targeting one pathway often prove ineffective for patients with different mutations. Professor Patrik Verstreken of the VIB-KU Leuven Center for Neuroscience emphasized that while patients share visible symptoms, molecular analysis reveals they fall into specific subcategories. Consequently, a one-size-fits-all drug for all Parkinson's variants does not currently exist. The research team adopted an unbiased approach to classify the disease. Instead of assuming how specific mutations affect the condition, they monitored fruit fly models carrying mutations in 24 different Parkinson's-related genes over time. First author Dr. Natalie Kaempf noted that the team entered the study without preconceived notions, allowing the data to guide the analysis. By applying computational and machine learning methods to this behavioral data, the researchers identified natural groupings that traditional hypothesis-driven methods would have missed. This strategy exposed a hidden structure within the disease, demonstrating that different genetic forms naturally cluster into distinct subtypes. The study also highlights the potential for clinical translation. Researchers found that animal models with specific subgroups could be cured by testing compounds tailored to those groups. When a drug that successfully treated one subgroup was tested on another, it failed to produce results. This confirms the possibility of developing subgroup-specific medications with precise positive effects. Dr. Verstreken explained that with these new categories, scientists can now search for specific biomarkers within each group and develop drugs tailored to their unique molecular dysfunctions. Beyond Parkinson's disease, this unbiased machine learning framework offers a new strategy for studying other complex conditions. The same principle could be applied to diseases caused by various gene mutations or environmental factors. By classifying such diseases based on underlying biological patterns rather than just symptoms, researchers can better understand their diversity and develop more effective, targeted interventions. The study stands as a testament to how artificial intelligence can reveal clinically meaningful variations in disease biology that remain invisible through conventional approaches.