On the heels of new research from Boston University showing that an artificial intelligence model was able to predict a person’s chances of developing Alzheimer’s disease, Weill Cornell Medicine researchers have been able to classify Parkinson’s disease into three subtypes using machine learning.
The findings — which appear in npj Digital Medicine — hold promise in aiding researchers and clinicians to target treatments specific to those subtypes.
Researchers at Cornell analyzed data from 406 people who participated in the Parkinson’s Progression Markers Initiative (PPMI), which is an international observational study that “systematically collected clinical, biospecimen, multi-omics, and brain imaging data of participants.”
They developed a deep-learning model called deep phenotypic progression embedding (DPPE), which was able to “holistically” model “multidimensional, longitudinal progression data of the participants,” as the authors explain in the study paper.
The authors further note that in recent years there has been a move toward observing Parkinson’s as a condition with heterogeneous symptoms and progression.
Not all individuals with Parkinson’s will have the same experience, in other words, and therefore treatment could be much more tailored to suit different patients’ needs.
The study authors note that their classifications “highlighted the necessity of treating Parkinson’s disease subtypes as unique sub-disorders within clinical practice, where our pace subtypes could inform patient stratification and management.By identifying specific varieties of the disease, clinical approaches could be much more targeted and effective.The goal of precision medicine is to predict the disease course in a patient and to therapeutically intervene ahead of time to prevent complications from developing. To achieve this we need to identify the disease driver in each patient and develop targeted therapeutics.
For example, they have found that 10% of Parkinson’s patients in the United States have a mutation in the GBA gene and that different types of GBA mutations accelerate the course of the disease.Patients with GBA mutations can now be enrolled in clinical trials for targeted therapies and ultimately will benefit from disease-modifying GBA-directed therapeutics.Early intervention may be required for rapid progressive patients.
This is crucial for managing symptoms before they become severe and debilitating. Subtyping helps in stratifying patients based on their risk, enabling more focused and effective clinical trials for new treatments, as well as better allocation of healthcare resources.While AI models are powerful tools for identifying disease subtypes and predicting progression, there are potential issues related to patient access. Not all patients may have access to advanced diagnostic tools or treatments derived from AI research, especially in under-resourced settings.
However, according to him, another issue might be the use of extensive patient data for AI model training which raises concerns about data privacy and security.AI models need to be validated across diverse populations to ensure they are not biased towards specific cohorts.
Scherzer, echoing his earlier statement, said that the significant power of artificial intelligence toward precise medical treatments will ultimately depend on more research and trials.The success of AI to predict outcomes depends on the size and quality of the input data.A key gap in the field is that we need much larger, high quality, longitudinal data sets of Parkinson’s patients — data on large populations spanning prodromal stages and the entire disease course. These will be essential for training and validating AI models useful for augmented medicine.
