Multimodal machine learning study of neurodegenerative diseases
Neurodegenerative disease is a leading cause of death in the world. However, limited treatment options are currently available. The changes of preclinical symptoms could happen years if not decades earlier before the threshold of clinical diagnosis is met. It is therefore important to identify people of high risk as early as possible. Recent machine learning models are able to combine different data types for accurate prediction. The objective of this study is to leverage the huge amount of data collected from UK Biobank and other large cohorts, and build machine learning methods to predict disease risk and subtypes. Our study would have the potential to identify people of high disease risk at an early stage, which would help future clinical intervention and risk mitigation. The study is anticipated to last for five years.
The objective of this study is to develop machine learning methods to integrate multimodality data to obtain a more comprehensive picture of neurodegenerative disease risk and subtypes. Two aims will be pursued: 1) To develop supervised learning models to predict disease risk; 2) To develop unsupervised learning models to identify disease subtypes.
In addition to neurodegenerative diseases, we will also apply a similar methodology to develop both supervised and unsupervised multimodal machine learning models to study the risk and heterogeneity of biological aging and cardiovascular disease, and their contribution to neurodegenerative diseases. Our hypothesis is that biological aging and cardiovascular disease are important factors associated with an increased risk of neurodegenerative diseases (especially vascular dementia). By building machine learning models, we might be able to better understand the relative contribution of biological aging and cardiovascular disease on neurodegenerative diseases compared to standard risk factors.