Precision nutrition (PN) is an emerging field that aims to provide the right diet to the right individual at the right time, which has the potential to improve diet-related prevention of several cardiometabolic diseases. PN can be provided on a personal level or a group level where individuals with similar metabolic regulation, i.e., metabolic phenotypes(metabotypes), are given the same dietary recommendations since they would likely respond similarly to the diets. To assess the metabolic response to a diet, metabolites derived from the metabolism, external exposures, and their interactions are often measured along with clinical markers such as glucose, insulin, and hormonal levels in the blood. Different metabotypes have been identified using such measurements and differential associations to disease outcomes and dietary challenges have been demonstrated, which highlights the importance of using them as a population-based approach to PN. In PN, two important challenges are considered; i) to predict individual postprandial metabolic responses to infer the best possible diet to the individual’s needs, and ii) to identify metabotypes to which diet may be tailored for improved preventive efficacy. Prediction of postprandial metabolic responses has yet to be conducted in a few disease-related features in repeated samples. Metabotyping has been performed using cluster analysis on data from static blood markers or from responses to single dietary challenges. However, methods incorporating data from several dietary challenges, or the postprandial dynamics have not been explored properly.
This thesis focuses on the development of methodologies to advance the progress of PN, specifically to address the two important challenges of prediction of postprandial response and to identify metabotypes using metabolomics, gut microbiota, dietary- and health status data. Rank-reduced linear dynamical systems derived using Dynamic Mode Decomposition were used to investigate the predictability of postprandial metabolic response using the baseline metabolome and nutritional information of meals. The method was shown to be predictive and interpretable in both measured (R2=0.4) and simulated (R2=0.65) data. Furthermore, it was also used along with the tensor decomposition CANDECOMP/PARAFAC to investigate methodologies to perform metabotyping on crossover intervention studies using repeated measures from several dietary challenges. Both methods successfully identified the same two metabotypes relating to amino acid uptake, although the method using tensor decomposition was shown to be more intuitive. Clustering of kinetic model parameters derived from postprandial plasma glucose dynamics was investigated to identify differential responders to meal challenges. Identified clusters associated differently with type-2 diabetes risk markers and gut microbiota, which showed that differences in postprandial dynamics relate to type 2 diabetes risk markers and can be used to identify individuals at risk. Finally, we investigated if metabotypes could be identified using metabolomics, microbiome, and cardiometabolic health profiles in free-living individuals (n=720) sampled three times during 1 year. Metabotypes related to metabolic syndrome were identified, which were well described by both metabolome and microbiome.
In summary, the developed analytical methodologies have provided a new toolbox to aid the advancement of PN in terms of metabotyping using data from more complex study designs, enabling dynamic predictions of postprandial response to food and showcasing that postprandial dynamics can aid in detecting individuals at disease risk.
Tuesday 23 January, 10:00 am