Construction and Application of Comprehensive Knowledge Graphs for Alzheimer's Disease

Univ Of North Carolina Chapel Hill, Chapel Hill NC

Investigators

Abstract

Project Summary/Abstract Despite significant advances in omics studies and laboratory observations, identifying the causal mechanisms for late-on- set Alzheimer's disease (AD) and AD-related dementia (ADRD) remains a challenge due to their multifactorial inheri- tance, which is heterogeneous across subjects and populations. In dementia studies, only a subset of relevant data is typi- cally collected from each participant, resulting in a biased exploration of AD pathogenesis. To address these limitations and respond to the NOT-AG-21-045 call for proposals on harmonizing complex data sets relevant to AD/ADRD, we pro- pose the development of a comprehensive AD-related knowledge graph (AD-KG) platform. This platform will inte- grate and harmonize data from multiple curated resources, literature, and a large-scale AD-related database with multiple types of high-dimensional data, including imaging, genetics, and clinical variables, from different studies with different missing-data patterns. Our inspiration for this proposal comes from the success of knowledge graphs (KGs) as a foundation for cognitive systems in industry, such as Microsoft XiaoIce. We propose to achieve four aims: Aim 1: Con- struct a dynamic AD-KG by utilizing multiple curated resources, literature, and individual data across different domains and studies for AD. Aim 2: Develop an omics knowledge graph (OKG) platform for harmonizing, imputing, and repre- senting AD-related multi-omics data. Aim 3: Develop a neuroimaging knowledge graph (NKG) platform for presenting, imputing, and representing AD-related neuroimaging data and a neuroimaging omics knowledge graph (NOKG) platform for neuroimaging-omics association maps. Aim 4: Analyze data from the large-scale AD-related database and verify the effectiveness of the newly developed AD-KG for clinically transformative research. The construction of OKG, NKG, and NOKG will complement the development of AD-KG. By developing a comprehensive AD-KG platform that integrates diverse data sources and types, we can facilitate more comprehensive research in AD and ADRD, leading to the develop- ment of new diagnostic and therapeutic approaches. Achieving these aims will offer unique analytic and data science ca- pabilities necessary for AD-related cognitive systems and greatly enhance cognitive techniques through innovative use of semantics and graphs to address complex data modeling, blending, and analytic challenges. To achieve our proposal aims, we have formed a team of experts in cognitive systems, knowledge graphs, statistical genetics/genomics, AD genetics, neuroimaging analysis, neuroscience, and statistics. Clinically, achieving these aims will enhance the identification of new genes and genetic pathways, leading to risk and protective factors for AD and inspiring novel therapeutic approaches. We plan to share our AD-KG, new cognitive tools, and AD-KG-based structured data with the research community through NIAGADS and other NIA infrastructure.