Innovative Machine and Deep Learning Analyses of Alzheimer's Disease Omics and Phenotypic Data

AD is the most common cause of dementia and presents a substantial and increasing economic and social burden. Our ability to diagnose and classify AD from cognitive normals (CN), or discriminate among individuals with AD, early mild cognitive impairment [EMCI], or late mild cognitive impairment (LMCI), is essential for the prevention, diagnosis, and treatment of AD. Since individuals with MCI have a high chance of converting to AD, effectively discriminating between those who convert to AD (MCI-C) from those who do not convert (MCINC) is important for early diagnosis of AD. The heterogeneity of AD has motivated attempts to classify distinct subgroups of AD to better inform the underlying physiology. There is evidence to suggest that using data across multiple modalities (e.g. genetics, imaging, metabolomics) has potential to classify AD subgroups better than using single modality. We will apply machine and deep learning methods to gain deeper insight into AD and ADRD pathobiology. We will use datasets that include genomics, genetics, metabolomics, and phenotypic data for this purpose. Data will be divided into discovery and validation sets. On the discovery set, state-of-the-art ML and DL methods for integrative analysis that we and others have developed will be coupled with resampling techniques to determine candidate molecular signatures and pathways discriminating the AD groups considered. Molecular scores will be developed from these candidate biomarkers. The clinical utility of the scores beyond well-known clinical risk factors for AD will be ascertained. We will validate our findings using the validation data. We will visually and quantitatively compare the risk scores across several clinical variables and outcomes. We will use (un)supervised clustering methods to identify molecular clusters, and we will investigate molecular clusters differentiating MCI to AD converters from non-converters. We may explore differences across ethnic subgroups. We will also innovatively apply our multimodal molecular subtyping methods to discover, reproduce, and characterize novel molecular subgroups of AD– this will allow for better risk stratification.

We have been developing novel machine learning (ML) and deep learning (DL) methods that leverage genomics, other omics (including proteomics and metabolomics), clinical and epidemiology data to better understand the pathogenesis of complex diseases. By integrating data from different sources, we have identified molecular signatures contributing to the risk of the development of complex diseases beyond established risk factors. We are proposing to innovatively apply these, and other existing, methods, to data pertaining to Alzheimer’s disease (AD) and Alzheimer’s disease related dementias (ADRD). A deeper understanding of the genes, genetic pathways, and other molecular signatures of AD is essential and could facilitate the identification of potential therapeutic targets for the disease.