Developments in personalised approaches to nutrition are hindered by a lack of large-scale high resolution data integrating multiple dietary, lifestyle, physiological, genetic and metagenomic data. Understanding the role of the integrated multiple regulatory systems involved during the dynamic postprandial phase is key to unravelling inter-individual variations within a healthy phenotype and advancing precision nutrition. In the largest and most detailed studies of metabolic responses to food, the Personalized Responses to Dietary Composition studies (PREDICT) assesses the genetic, metagenomic, metabolomic and meal-context drivers of postprandial metabolic responses to predict individual responses to food using a machine learning algorithm.
The PREDICT 1 multi-center postprandial study of 1,000 individuals from the UK (unrelated, identical and non-identical twins) and 100 unrelated individuals from the US, assessed postprandial (0-6h) metabolic responses to sequential mixed-nutrient dietary challenges in a clinic setting. Glycemic and lipaemic responses to multiple duplicate isocaloric meals of different macronutrient content and self-selected meals (>100,000), were tested at home using a continuous glucose monitor (CGM) and dry blood spots. Baseline factors included metabolomics, genomics, gut metagenomics and body composition. Dietary data was collected using the Zoe app and dashboard combining weighed food records, photographs, bar coding and live nutritional support. Sleep and activity were monitored using wearable devices. The ‘Big Data’ collected in PREDICT included 2 million glucose responses, 28,000 triacylglycerol (TG) measurements and 132,000 weighed meal logs.
Inter-individual variability in postprandial responses (glucose, insulin and TG) was high in the clinic setting, even between identical twins. Genetic contributions, determined by classical twin methods, for glucose, insulin and TG responses was less than 50%, 30% and 5% respectively. Identical twins shared 37% of their gut species, compared to 35% for unrelated individuals. Interim machine learning algorithm predicted 77% of the variation in glycemic responses. Only approximately 29% of variation could be explained by the proportion of macronutrients in the meal.
Meal context, including time of day, exercise and meal sequence impacted postprandial glycaemic responses. Given that traditional measures of postprandial glycaemic and liapemic responses (iAUC) do not reflect all features of postprandial responses which impact health, different features of the postprandial response curve were explored. For example, the 2-3 hour glucose dip below baseline had a large impact on satiety, hunger and energy intake, whereas the 2h iAUC had only minimal impact. Individual’s experiencing greater glucose ‘dips’, experienced lower levels of satiety and consumed more energy at the subsequent meal and over the following 24 hour period. The 6h rise in plasma TG was the mostly closely associated feature of the postprandial lipaemic response curve with atherogenic lipoproteins and inflammatory measures. Postprandial induced inflammation (IL-6 and GlycA) was highly variable between individuals. Postprandial IL-6 was only weakly correlated with postprandial glycaemia and lipaemia. However postprandial GlycA was strongly correlated with lipaemia (r=0.832), suggesting that postprandial induced inflammation is predominantly lipid mediated.
The gut microbiome is shaped by diet and influences host metabolism, but these links are complex and can be unique to each individual. We therefore performed deep metagenomic sequencing of PREDICT participants. We found strong associations between microbes and specific nutrients, foods, food groups, and general dietary indices, driven especially by the presence and diversity of healthy and plant-based foods. Microbial biomarkers of obesity were reproducible across cohorts, and blood markers of cardiovascular disease and impaired glucose tolerance were more strongly associated with microbiome structure. While some microbes such as Prevotella copri and Blastocystis spp., were indicators of reduced postprandial glucose metabolism, several species were more directly predictive for postprandial triglycerides and C-peptide. The panel of intestinal species associated with healthy dietary habits overlapped with those associated with favourable cardiometabolic and postprandial markers, indicating our large-scale resource can potentially stratify the gut microbiome into generalizable health levels among individuals without clinically manifest disease.
The large and potentially modifiable variation in metabolic responses to identical meals in healthy people explains why ‘one size fits all’ nutritional guidelines are problematic. The ongoing PREDICT programme of research, collecting more information on glucose, TG and multi-omic responses to thousands of meals, alongside environmental, genetic, clinical and microbiome variables, give excellent power to use machine learning to optimise and predict individual responses to foods and provide personalised nutrition advice.