Nutritional metabolomics: capturing habitual diet patterns and their relation to health outcomes in the population.

The program goal is to use nutritional metabolomics to objectively describe intake of major food groups in healthy diets, combined with self-reported intake, among individuals in large epidemiological cohorts and link the intake to later health.

This project aim is to discover putative biomarkers for different protein sources, for berries, fruit and vegetables and for whole grain. A sample of 210 adults will be extensively characterized on habitual diet and phenotype. Of these 60 will be included in a cross-over feeding trial.

Prediction models of trial diets will be developed from metabolite patterns in blood and urine using machine learning and validated in data on habitual intake from the 210 adults and another study.

The models will be applied to a subsample of the Västerbotten cohort with metabolomics data (n=4000) to calibrate self-reported intake data. Dietary intake data from the entire cohort (n=120 000) will then be linked to later mortality and morbidity.

Winkvist, Lindqvist and Ellegård will during 2025-2027 run the project at Gothenburg Univ which houses facilities for feeding trials. Machine learning, conducted with Chalmers, statistical analyses and publications will take place during 2027-2028. Globally sub-optimal diet is a leading risk factor for premature death.

Dietary intake is difficult to measure, hampering research and preventive measures. Project results will help monitoring adherence to and effects of the new national guidelines for a healthy diet.