Dissertations and Theses

Date of Degree

6-3-2020

Document Type

Dissertation

Degree Name

Doctor of Philosophy (Ph.D.)

Department

Epidemiology and Biostatistics

Advisor(s)

Mary Schooling

Committee Members

Marlene Camacho-Rivera

Shiro Horiuchi

Subject Categories

Public Health

Keywords

Alzheimer's Disease, metabolites, Mendelian randomization

Abstract

This paper uses two sample Mendelian randomizations to conduct an agnostic search of the metabolic GWAS published by Tao Long and co-workers. GWAS for AD from IGAP and from UK Biobank were used for outcomes. The author uses two sample Mendelian randomization (MR) to test for causal relationships between common blood metabolites and Alzheimer’s disease (AD). First, estimates for a set of 22 metabolites which are statistically significant for both outcome GWAS are determined. Egger regression, Mendelian Randomization Pleiotropy Residual Sum and Outlier (MRPRESSO), and weighted median regression are used as sensitivity checks. AD shares risk factors with both heart disease and cancer, and it is possible that individuals highly susceptible to AD may have died of a competing risk before recruitment into the study. To examine this further, two sample MR is used to determine if there are causal relationships the 22 metabolites and AD. Relationships are found for heart disease in 17 of the 22 metabolites, and in 2 of the metabolites for cancer. Findings strongly suggested competing risk bias in the aggregate, at least for heart disease. To determine how this general finding applied to metabolites under consideration, estimates for AD were obtained by sex. Women have a higher rate of all-cause survival than men, so any survival bias present should cause the estimates for men to be attenuated. Therefore, the UK Biobank proxy phenotype GWAS was examined by sex and the author found 15 metabolites not affected by competing risk bias. Among the blood metabolites that were examined, the best candidates for causal relationship to Alzheimer’s disease were 1-stearoyl-2-arachidonoyl-GPE (18:0/20:4), cis-4-decenoyl carnitine, and 1-palmitoyl-2-linoleoyl-GPE (16:0/18:2).

Available for download on Saturday, May 20, 2023

Included in

Public Health Commons

Share

COinS