Identification of Metabolic Phenotypes and Systemic Biochemical Reaction Networks Associated with Human Blood Pressure

Lead Research Organisation: Imperial College London
Department Name: School of Public Health


High blood pressure is an important component of metabolic syndrome and one of the most common modifiable risk factors for human cardiovascular disease and non-communicable disease. High blood pressure remains to be the leading cause for global disease burden and is estimated to attribute to 10.7 million deaths globally.
Different clinical and population-wide strategies have been in use to reduce high blood pressure such as to slow the rise of obesity, to increase physical activity, to reduce tobacco use and to change dietary patterns including the reduction of salt intake.
However, better knowledge of the effects of high blood pressure on human metabolism is needed to provide insight into the molecular mechanisms by which blood pressure is controlled. The human metabolic phenotype reflects the health status of an individual and is affected by genetic, environmental, dietary and gut-microbial variation.
I propose to use a metabolic phenotyping approach to discover biochemical compounds (metabolites) in urine related to blood pressure and associated dietary patterns, and to investigate how these potential clinical metabolic targets fit in the biochemical processes involved with the onset of high blood pressure and homeostatic regulation of blood pressure.

Technical Summary

I aim to evaluate the relationship between the urinary metabolic profile and blood pressure, including the linkage with adiposity, salt intake and associated dietary patterns and map these in a holistic systems framework.
A) To uncover the metabolic signature of blood pressure by means of a Metabolome-Wide Association Study with control for confounding variables using 1H Nuclear Magnetic Resonance (NMR) spectroscopy and Direct Injection Mass Spectrometry (DIMS) data.
B) To identify novel associations between the intake of dietary nutrients and excretion of urinary metabolites to pinpoint dietary patterns associated with cardiometabolic risk.
C) To investigate the interdependence between cation (calcium, magnesium, potassium and sodium) and metabolite excretions with respect to renal function and (re-)absorption, and homeostatic regulation.
D) To map the pathways perturbed by high blood pressure, sodium consumption and dietary patterns, and assess the overlap with the metabolic and systemic signature of adiposity and other risk factors to reveal the integrated systems network of blood pressure.
Methodology: I propose to use the Metabolome-Wide Association Study paradigm on 24-hour urine samples measured with 1H NMR spectroscopy and DIMS, and enhance this by coupling multivariate statistical methods with the analysis of integrated systems metabolic reaction networks.
Outcomes: This work will improve the understanding of the molecular mechanisms underlying high blood pressure and define new potential starting points for the development of appropriate and comprehensive preventive and treatment strategies to address the increasing problem of blood pressure worldwide.


10 25 50
Description School Engagement Activity - St Andrews University (HDR-UK Summer School): Man vs Machine Learning 
Form Of Engagement Activity Participation in an activity, workshop or similar
Part Of Official Scheme? No
Geographic Reach Regional
Primary Audience Schools
Results and Impact 45 secondary school children (15-16 years old, from two schools) took part in an engagement activity to learn about pattern recognition and machine learning on Friday the 23rd of August at the University of St Andrews, Scotland. They were first introduced to heart physiology and anatomy by local GP and researcher Dr David Fraile Navarro. With this background knowledge they then were shown the activity by Dr Joram Posma where groups of 2-3 students each would compare electrocardiograms (ECGs) from six healthy patients with those from patients that have had a myocardial infarction (also six). Each group was then given 8 minutes to classify as many as they could from a set of 26 ECGs to either healthy or myocardial infarction by trying to spot differences between healthy and myocardial infarction ECGs. The aim of the exercise was to get the students to apply human pattern recognition system to these ECGs. After the 8 minutes were up the results were collected and compared against the true classification and a simple machine learning algorithm that classified the same unknown ECGs with the same training data. We then explained the difference between false positives and false negatives and what these mean in terms of classification of healthy and ill individuals and its importance for treatment options. The activity was rounded off by comparing the aggregated results of each group against the machine learning algorithm and discussing the current and future use of machine learning algorithms in routine healthcare. With the limited knowledge the students had of ECGs they still managed to correctly classify 61.5-88.5% of ECGs, compared to 92.3% of the machine learning algorithm, with 7.7-33.3% of false positives and 15.4-42.9% false negatives (compared to 0% false positives and 14.3% of false negatives of the machine learning algorithm). This highlighted both the benefit of machine learning as well as the improvements new health data scientists can make to improve the accuracy and limit false negatives.
During lunchtime the students were asking how I used machine learning in my day-to-day research which allowed me to explain how I use different methods to identify compounds in urine that relate to differences in blood pressure. The feedback received from teachers and students was very positive, teachers reported increased interest in (health) data science and students considered data science as a career choice (some previously only considered medicine or biology as means to 'help save lives').
Year(s) Of Engagement Activity 2019