Metabolomic-based identification of clusters that reflect dietary patterns

Scope: Classification of subjects into dietary patterns generally relies on self-reporting dietary data which are prone to error. The aim of the present study was to develop a model for objective classification of people into dietary patterns based on metabolomic data. Methods and results: Dietary and urinary metabolomic data from the National Adult Nutrition Survey (NANS) was used in the analysis (n=567). Two-step cluster analysis was applied to the urinary data to identify clusters. The subsequent model was used in an independent cohort to classify people into dietary patterns. Two distinct dietary patterns were identified. Cluster 1 was characterized by significantly higher intakes of breakfast cereals, low fat and skimmed milks, potatoes, fruit and fish, fish dishes (P<0.05) representing a 'healthy' cluster. Cluster 2 had significantly higher intakes of chips/processed potatoes, meat products, savory snacks and high-energy beverages (P<0.05) representing an 'unhealthy cluster'. Classification was supported by significant differences in nutrient status (P<0.05). Validation in an independent group revealed that 94% of subjects were correctly classified. Conclusion: The model developed was capable of classifying individuals into dietary patterns based on metabolomics data. Future applications of this approach could be developed for rapid and objective assignment of subjects into dietary patterns.