Abstract
Our overarching goal is to construct heritability enhanced multimodal imaging endophenotypes, dissect their
associated genetic underpinnings, and characterize the pathological mechanism along with clinical profiles for
Alzheimer’s disease (AD) susceptibility and progression through a suite of statistically powerful, biologically
plausible and computationally efficient Bayesian models. AD is complex: its etiology is in no way clear and with
no available cure. Genetics play a dominate role in AD. However, genome wide association studies based on
AD susceptibility have only discovered a small number of independent genetic factors. Compared with
categorical diagnoses, imaging quantitative trait (QT) has distinct advantages to capture disease etiology and
their association on genetic variants has yielded some prominent finding. However, existing imaging genetics
studies suffer with i) inadequate power under current atlas-based imaging traits, ii) oversight of genetic
underpinnings across different brain processes and potential role of SNP-SNP interactions, and iii) unclear
pathological mechanism among genetics, imaging and clinical profiles, which limit their further advances to AD
research. To overcome these barriers, this project proposes the following four aims: 1) construct novel
multimodal imaging endophenotypes under “brain heritability parcellation'” for AD risk and progression; 2) identify
pleiotropic SNPs and SNP-SNP interactions under constructed multimodal brain biomarkers; 3) investigate the
fundamental pathological mechanism between genetic variations and AD pathology mediated by multimodal
imaging endophenotypes; and 4) perform systematic evaluation of the proposed methods through real data
analyses and simulations, and develop user-friendly analytical pipelines. Our proposed methods are innovative
in multiple aspects for and beyond AD biomedical research including a) to construct a novel brain heritability
parcellation, b) to integrate multimodal imaging on their association to genetic bases, c) to explore risk SNP-
SNP interaction along with common genetic variants in a coherent and scalable way, d) to consider genetic
pleiotrophy in a fine mapping framework, e) to establish pathological mechanism among individual’s genetic
variants and polygenic profile, brain activities and disease symptoms, f) to extend the proposed methods to
longitudinal settings with respect to AD progression, and g) to develop efficient and user friendly pipelines for
our products. A successful completion of this proposal will facilitate the identification of novel genetic variants
and characterize their impacts on AD pathogenesis, mediated by the constructed multimodal imaging biomarkers.
The development of innovative statistical methods and computational tools, and their implementation on the
ADNI and Yale ADRC cohort will offer a great potential for understanding the genetic and neurobiological
mechanisms of AD and advancing targeted prevention and treatment, paving the way for the development of
neurological and psychiatric research in general and benefit public health outcomes.