Machine Learning Reveals Three Subtypes of Parkinson’s Disease Progression

Researchers at Weill Cornell Medicine have made a significant breakthrough in understanding Parkinson’s disease by using machine learning to define three distinct subtypes based on the pace of disease progression. Led by Dr. Fei Wang, a professor of population health sciences and director of the Institute of AI for Digital Health, the team analyzed de-identified clinical records from two large databases and identified specific driver genes associated with each subtype.

The subtypes, named Inching Pace, Moderate Pace, and Rapid Pace, are characterized by distinct patterns of disease progression and molecular mechanisms. For instance, the Rapid Pace subtype is marked by activation of pathways related to neuroinflammation, oxidative stress, and metabolism. The researchers also identified potential drug candidates that could be repurposed to target specific molecular changes seen in each subtype. Notably, they found that patients taking the diabetes drug metformin appeared to have improved disease symptoms, particularly those with cognitive deficits.

This research has significant implications for developing customized treatment strategies for Parkinson’s patients and highlights the potential of machine learning in advancing our understanding of complex diseases.

Machine Learning Helps Define New Subtypes of Parkinson’s Disease

Researchers at Weill Cornell Medicine have utilized machine learning to identify three subtypes of Parkinson’s disease based on the pace at which the disease progresses. This breakthrough has the potential to become an essential diagnostic and prognostic tool, as well as suggest ways to target these subtypes with new and existing drugs.

The study, published in npj Digital Medicine, employed deep learning-based approaches to analyze de-identified clinical records from two large databases. The researchers also explored the molecular mechanisms associated with each subtype through the analysis of patient genetic and transcriptomic profiles using network-based methods. This led to the identification of distinct brain imaging and cerebrospinal fluid biomarkers for the three subtypes.

The three subtypes, named Inching Pace (PD-I), Moderate Pace (PD-M), and Rapid Pace (PD-R), were defined based on their distinct patterns of disease progression. PD-I is characterized by mild baseline severity and slow progression speed, affecting approximately 36% of patients. PD-M has mild baseline severity but advances at a moderate rate, accounting for around 51% of cases. PD-R is marked by the most rapid symptom progression rate.

Distinct Molecular Mechanisms Associated with Each Subtype

The researchers found that each subtype is associated with distinct molecular mechanisms. For instance, the PD-R subtype had activation of specific pathways related to neuroinflammation, oxidative stress, and metabolism. These findings have significant implications for the development of targeted therapeutic strategies.

Furthermore, the study identified possible drug candidates that could be repurposed to target the specific molecular changes seen in the different subtypes. The researchers used two large-scale, real-world databases of patient health records to confirm that these drugs could help ameliorate Parkinson’s progression.

Potential Therapeutic Applications

One promising finding was that people taking the diabetes drug metformin appeared to have improved disease symptoms, particularly those related to cognition and falls, compared with those who did not take metformin. This was especially true in individuals with the PD-R subtype, who are most likely to experience cognitive deficits early in the course of their Parkinson’s disease.

The study’s results suggest that machine learning can be a powerful tool for identifying new therapeutic opportunities in Parkinson’s disease. By examining diverse data sources and leveraging advanced computational methods, researchers may uncover novel ways to target this complex and debilitating condition.

Collaborative Efforts and Future Directions

This work was made possible through collaborative efforts involving scientists from multiple institutions, including the Cleveland Clinic, Temple University, University of Florida, University of California at Irvine, and University of Texas at Arlington. The study’s findings have significant implications for the field of translational bioinformatics, and future research directions may involve further validating these results both computationally and experimentally.

The institution’s commitment to transparency is reflected in its public disclosure of relationships between Weill Cornell Medicine physicians and scientists with external organizations. This work was supported by grants from the National Institute on Aging, the National Institute of General Medical Sciences, and the National Institute of Neurological Disorders and Stroke, all part of the National Institutes of Health.