Study Rationale:
Parkinson's disease is highly variable, so each patient has a unique experience of the disease and diverse contributing factors. To understand and ultimately treat Parkinson's, we need to identify those differences as disease subtypes. The goal of this project is to develop data-driven approaches for identifying cohesive subtypes for Parkinson's patients in BioFIND study using their clinical and genetic information.
Hypothesis:
Our hypothesis is that the patients within each subtype will demonstrate similar patterns in their clinical records and genetic information.
Study Design:
We will utilize de-identified patient information, including clinical and genetic data, from the BioFIND study. We will leverage state-of-the-art machine learning techniques including deep learning, clustering and visualization to identify patient clusters. Patients should have high intra-cluster similarities comparing to inter-cluster ones. Each patient cluster will be compared with classic Parkinson's subtypes as well as the subtypes derived from our previous study using data from the Parkinson's Progression Markers Initiative (PPMI).
Impact on Diagnosis/Treatment of Parkinson's disease:
Identifying Parkinson's disease subtypes can help us better understand the variability of the disease. This can help researchers and clinicians develop better therapeutics and pave the way for precision medicine approaches and appropriate therapeutic plans.
Next Steps for Development:
Our next step is to check the agreement between the subtypes derived from BioFIND and PPMI and then attempt to replicate the subtypes in more Parkinson's cohorts.