Metabolic variation reflects dietary exposure in a multi-ethnic Asian population

Understanding how diet shapes metabolism across diverse populations is essential to improving nutrition and health. Biomarkers reflecting diet are explored largely in European and American populations, but the food metabolome is highly complex and varies across region and culture. We assessed 1,055 plasma metabolites and 169 foods/beverages in 8,391 multi-ethnic Asian individuals and carried out diet–metabolite association analyses. Using machine learning, we developed multi-biomarker panels and composite scores for key foods, beverages and overall diet quality. Here we show these biomarker panels can be used to objectively assess dietary intakes in the Asian multi-ethnic population and can explain variances in intake prediction models better than single biomarkers. The identified diet–metabolite relationships are reproducible over time and improve prediction of clinical outcomes (insulin resistance, diabetes, body mass index, carotid intima-media thickness and hypertension), compared to self-reports. Our findings show insights into multi-ethnic diet-related metabolic variations and an opportunity to link exposure to population health outcomes. In a large multi-ethnic Asian cohort, associations between over 1,000 plasma metabolites and specific foods and beverages are made. These diet–metabolite relationships were used to accurately predict clinical phenotypes such as diabetes and hypertension.