mattlee821 / 000_thesis

PhD thesis

PhD thesis in Population Health Science

What lies behind the causal impact of body mass index level and change on human health? Added value from complementary study design and deep metabolomic phenotyping

Supervisors: Nicholas Timpson, Kaitlin Wade, Laura Corbin

This GitHub repository houses all of the components used in the creation of my thesis. Importantly, it makes available all publicly available data, scripts, results, and figures used in the thesis. All figures presented in the thesis, and those too large to be included, are available within this repository. The structure of the repository is as follows:

1. All data, results, figures, and scripts mentioned in the thesis are within the index\ directory
2. Within index/:
   • _book/ - contains the thesis as a complete PDF (thesis.pdf) as well as each individual chapter (all are accompanied by the .tex files
   • bib/ - contains the bibtex file used to create the references
   • csl/ - contains the citation style language (csl) file for references
   • data/ - contains all chapter specific scripts, data, results, and figures
   • .Rmd - these are the R Markdown files used to write components of the thesis
   • _bookdown.yml - yml file used to compile the thesis
   • bristolthesis.cls - class file used by template.tex
   • template.tex - LaTeX template file for creating thesis
3. Within index/data/ each chapter has a separate directory with a similar file structure:
   • scripts/ - all scripts used in the chapter
   • data/ - all data used in the chapter
   • analysis/ - all products, primarily results, from scripts
   • figures/ - all figures used in chapter
   • tables/ - tables presented in or referenced in the thesis

The final version of the thesis can be found in index/_book. GitHub does not always render large PDFs so you will probably need to download to read. The thesis is split into the following chapters:

• Chapter 1 - Introduction: This chapter gives the background to the thesis and details the aim and objectives. All data and scripts used in this chapter is available in index/data/introduction
• Chapter 2 - Systematic review and meta-analyses: This chapter details a systematic review and meta-analyses of all Mendelian randomization analyses which use adiposity measures as an exposure. All data, scripts and results for this chapter are available in index/data/SR
• Chapter 3 - Visualisation: This chapter details the development of EpiViz, an R package and web application for the production of Circos plots. All scripts and plots used in this chapter are available in index/data/visualisation
• Chapter 4 - Observational analyses: This chapter details a series of linear models using data from the Avon Longitudinal Study of Parents and Children of the effect of adiposity on metabolites. All scripts and results for this chapter are available in index/data/observational. The raw data for this work is not publicly available.
• Chapter 5 - Mendelian randomization analyses: This chapter details a series of two-sample Mendelian randomization analyses using publicly available data of the effect of adiposity on metabolites. All data, scripts, and results for this chapter are available in index/data/MR.
• Chapter 6 - Multivariable Mendelian randomization analysis: This chapter details a multivariable analysis using publicly available data to investigate the potential intermediary role of metabolites on the effect of adiposity on endometrial cancer. All data, scripts, and results for this chapter are available in index/data/mediation.
• Chapter 7 - Discussion: This chapter draws together all of the work in the preceding chapters to discuss the overarching themes of the thesis.
• Appendix: The appendix is split by each chapter
• References: a .bibtex file of the reference library is available in index/bib. References are formatted using the Nature style available in index/csl.

Abstract

Excess adipose tissue, adiposity, is associated with many diseases and overall mortality. This thesis aimed to investigate whether metabolites play an intermediary role in these associations using a variety of resources and methods to strengthen causal inference.

In a comprehensive systematic review and meta-analysis (Chapter 2), the causal effects of adiposity were observed across a broad spectrum of diseases, including endometrial cancer, which was selected for subsequent analysis. These results were supported by a narrative synthesis of over 2,000 Mendelian randomization (MR) analyses, which also highlighted evidence of an association between adiposity and many, predominantly lipid-based, metabolites.

Within the Avon Longitudinal Study of Parents and Children (ALSPAC), evidence for a consistent association between body mass index (BMI), waist hip ratio (WHR), and body fat percentage (BF) with up-to 230, predominantly lipid-based, metabolites and ratios was found (Chapter 4). In these linear models, the effect of adiposity persisted after adjustment for covariables and across the lifecourse.

In two independent datasets, MR, a method that mitigates limitations in observational analyses, provided further evidence for an association between BMI and WHR with up-to 230, predominantly lipid-based, metabolites and ratios (Chapter 5). These effects were consistent in sensitivity analyses. The effect of BF on metabolites was frequently opposite to the effects observed for BMI and WHR, and BF in observational analyses.

Evidence from observational and MR analyses identified 54 metabolites that were consistently associated with adiposity. Two of these (triglycerides in small and very small very large low density lipoprotein) were associated with endometrial cancer in MR analysis and, using multivariable MR, there was evidence for a potential intermediary role of both metabolites on the effect of WHR and BF, but not BMI, on non-endometrioid cancer, but not endometrioid cancer (Chapter 6). Weak instruments may have biased these results however.

This work highlights the broad effect of adiposity on the metabolome, identifies two metabolites that may be involved in the association between adiposity and endometrial cancer, and provides a basis for future investigations of the intermediary role of metabolites.

How

Thesis written in R Markdown using bristolthesis, an altered version of thesisdown that complies with the University of Bristol’s regulations. All formatting is controlled by template.tex and bristolthesis.cls. The _bookdown.yml file is used to compile the individual .Rmd files alongside index.Rmd.

Change log

I made corrections after my Viva in Januray 2022. All corrections are logged in the commit 3f89f4d82f3f47663d1fe4d93cd2f658db866a34 and as a response in examiner_corrections.docx. A final release was created after approval of these corrections (3 days after the commit).