Author

SHUDI LI

Date of Award

Fall 8-2020

Degree Name

Doctor of Philosophy (PhD)

Advisor(s)

MOMIAO XIONG

Second Advisor

Han Chen

Third Advisor

Bing Yu

Abstract

Alzheimer's disease (AD) and heart failure (HF) are two complex diseases that are caused by the combination of genetic and epigenetic, environmental and other lifestyle factors. Understanding the relationships between genetic and epigenetic variants and other factors of such complex diseases could assist researchers discover disease mechanisms and develop targeted therapies. Much of the research in genetics/epigenetics studies regarding AD and heart diseases have been focused on association analysis. Many researchers have identified genetic/epigenetics variants and phenotypes that are significantly associated with disease pathology. While most of these studies utilize association analysis as the analytical platform, the signals identified by association studies can only explain a small proportion of the heritability of complex diseases and a large proportion of risk factors remain undiscovered, which is the limitation of genome- wide association studies (GWAS). In addition, the biological system usually functions in a systematic or causal way, thus causation analysis is key to uncover the risk mechanisms of complex diseases. The relationship between association and causation is that causation can be used to infer association, but the reverse cannot be guaranteed. Traditionally, the gold standard for causation analysis is using interventions in randomized controlled trials (RCT). However, RCT is not feasible for genetics/epigenetics data for either ethical or technical reasons. The major objective of this research is thus to propose methods to uncover the causal mechanisms between genetic/epigenetic factors and phenotypes such as environmental and lifestyle factors for complex diseases. First, I proposed a bivariate causal discovery method to uncover the pairwise causal relationships between factors. Second, I proposed a network analysis framework to construct the causal network among genetic/epigenetic variants and phenotypic factors. Finally, I applied the bivariate causal discovery method and causal network construction method to the two complex diseases: Alzheimer's disease (AD) and heart failure (HF) data. Simulations and applications results were discussed in the following sections.

COinS