Learning molecular fingerprints of foods to decode dietary intake

Abstract

Assessing dietary intake from biological samples provides critical objective insights into nutrition and health. We present a reference-based strategy using untargeted metabolomics to estimate relative dietary composition. The approach learns food-specific molecular ion features first - both annotated and unannotated - via supervised classification and discriminant analysis. These features then guide extraction of corresponding MS1 intensities from unknown samples, enabling proportional, ion-resolved dietary readouts. Tracking these signatures across thousands of public datasets revealed feces, urine, and blood/plasma as optimal biospecimens. Validation with NIST omnivore/vegan stool samples, controlled mouse feeding study, food reintroduction trial in Crohn's disease, and a Mediterranean diet intervention trial confirmed that ion-resolved readouts reflect known intake patterns. In rheumatoid arthritis data, dietary scores obtained from MS/MS signatures correlated with clinical outcomes. To facilitate adoption, we developed an easy-to-use web-based "food readout" app. This method complements traditional diet assessments and advances personalized nutrition and nutritional epidemiology.

Figure 1.. Identification and classification of nutritional dark metabolome for dietary readout.

(a) Untargeted metabolomics workflow for obtaining the dietary scores from unknown samples. (b) Relative distribution of dietary metabolites in food classes, metabolites from food sources were annotated using retention time and MS/MS spectral matching. Unannotated metabolites were associated with food categories using a discriminant analysis model. (c) Discriminant analysis of food metabolomics data and development of food-specific MS/MS reference library. Library matching for detection of nutritional dark metabolites in unknown samples, following detection, peak area can be used to obtain relative dietary scores for each food category

Figure 2.. Reference-based spectral matching to food MS/MS library predicts dietary intake.

(a) Detection and distribution of specific dietary metabolites in human organs across GNPS-MassIVE (purple). Plant-based metabolites are highlighted with green circles. Light shading indicates the number of unique MS/MS spectra detected, while dark shading represents total detection frequencies. (b) Experimental study design where mice were fed with an anti-inflammatory and western diet, and fecal samples were collected at the end of 1and 2-week. (c) Principal component analysis shows clear separation of mice groups that are on regular chow (baseline), anti-inflammatory, and western diet. (d) Dietary intake prediction using dietary scores show differences in distinct food categories in mice with ITIS vs western diet.

Figure 3.. Dietary scores predict dietary intake in human fecal samples.

(a) In NIST reference standards, stool samples were collected from people with self-reported habitual omnivore and Vegan dietary habits. (b) Principal component analysis shows distinct separation of vegan and omnivore samples based on dietary scores. (c) Dietary scores show differences in distinct food consumption patterns between vegan and omnivore samples.

Figure 4.. Mediterranean diet crossover study

(a) In a randomised mediterranean diet crossover study, individuals were assigned to four different dietary groups. (b) Boxplots comparing dietary scores for mushroom, beef, fig and vegetable for each dietary period.

Figure 5.. Dietary scores infer macronutrient intake patterns from food diaries.

(a) Correlation matrix showing associations between dietary scores and macronutrient intake variables (Pearson correlations, n=698). Color intensity indicates correlation strength, with statistical significance denoted by asterisks (*p<0.05, **p<0.01, ***p<0.001). (b) Scatter plots demonstrating relationships between dietary scores and macronutrient content of individual meals, with regression lines and 95% confidence intervals shown (n=698). (SFA = Saturated fatty acids, DF/1000kcal = dietary fiber/1000kcal)

Figure 6.. Dietary scores link foods with clinical symptoms in patients with Rheumatoid arthritis (RA).

Pearson correlations showed associations of clinical disease activity indices (DAS28-CRP, CDAI & SDAI), swollen joint count (SJC), Tender joint count (TJC), pain visual analog scale (VAS_MD) and C-reactive protein (CRP) level with dietary scores. Few foods (#) have previously been reported by the clinical trials/animal models to impact RA disease states, and similar observations are replicated by metabolomics-based dietary readout.