AbstractBlood pressure (BP) is a complex multifaceted phenotype, and highly predictive of early cardiovascular events and mortality. Previous studies have identified multiple physiological pathways and biomarkers involved in BP regulation. While these studies have improved our understanding of the mechanisms underlying BP and hypertension (high BP), they have been mostly conducted in the context of single-omic analyses and have not been able to account for the mutual dependencies of the multiple interconnected pathways.
This thesis aims to provide a greater understanding of the physiology underpinning BP regulation by identifying the cross-sectional and longitudinal biomarkers of BP and hypertension in the context of single-omics, and then using machine learning methods to integrate datasets to identify the most important contributors.
To this end, multiple cohorts, including TwinsUK, the Personalised Responses to Dietary Composition Trial 1 (ZOE PREDICT-1), the Consortium of METabolomics Studies, and the Qatar Biobank were used to explore the associations with diet (Chapters 4 & 5), metabolites (Chapter 6), immunoglobulin G (IgG) glycans (Chapter 7), and gut microbes (Chapter 8), using unidimensional analyses. In Chapter 9, machine learning methods were used to integrate high-throughput datasets, including omics, biochemical, and dietary data, to identify the key features underlying BP regulation while accounting for their mutual dependencies.
In doing so, the role of diet, particularly nutrients, was confirmed, with tryptophan, riboflavin, alcohol, lactose, and biotin having the greatest effect on BP. A negative association between adherence to the Dietary Approaches to Stop Hypertension (DASH) diet and hypertension was also identified, but report that this effect is mediated by BMI. The largest metabolome-wide association study of hypertension to date, spanning 44,306 multi-ethnic individuals, identified five novel metabolite biomarkers, including two bile acids, two glycerophospholipids, and ketoleucine, and confirm 27 previous hypertension-associated metabolites. Additionally, an IgG glycan score that is predictive of incident hypertension was constructed. Furthermore, significantly lower gut microbial diversity was observed in hypertensive cases when compared to non-hypertensive controls, suggesting that the gut microbiome may be a target for interventions to improve BP regulation. Finally, using machine learning to integrate environmental, dietary, genetics, metabolites, and clinical data, in context of one another, a panel of the top 50 features contributing to BP regulation was constructed. These include traditional risk factors and the metabolome. This biomarker panel of the top 50 features is early in the clinical translation runway but provides an incremental advance to prioritise certain metabolic pathways in conjunction with diet and biochemical pathways.
In conclusion, novel biomarkers and pathways of BP regulation were identified and the value of a more holistic approach (e.g., machine learning) to dissect the pathways underlying BP regulation highlighted. Future studies should focus on the identified pathways to derive actionable targets for lifestyle or pharmacotherapy interventions.
|Date of Award||1 Mar 2023|
|Supervisor||Sarah Berry (Supervisor) & Cristina Menni (Supervisor)|