Skip to main content
ARS Home » Pacific West Area » Davis, California » Western Human Nutrition Research Center » Diet, Microbiome and Immunity Research » Research » Publications at this Location » Publication #415481
Research Project: Effect of Diet on Gut Microbiome, Gastrointestinal Health, and Immune Function

Location: Diet, Microbiome and Immunity Research

Title: Machine learning with ingredient-level food trees reveals contributors to systemic inflammation among adults in the National Health and Nutrition Examination Survey, 2001-2010 and 2015-2018

Author
item Larke, Jules
item Lemay, Danielle

Submitted to: Journal of the Academy of Nutrition and Dietetics
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 5/22/2025
Publication Date: 5/28/2025
Citation: Larke, J.A., Lemay, D.G. 2025. Machine learning with ingredient-level food trees reveals contributors to systemic inflammation among adults in the National Health and Nutrition Examination Survey, 2001-2010 and 2015-2018. Journal of the Academy of Nutrition and Dietetics. https://doi.org/10.1016/j.jand.2025.05.012.
DOI: https://doi.org/10.1016/j.jand.2025.05.012

Interpretive Summary: The relationship between diet and inflammation is complex and difficult to accurately quantify from dietary recalls. In this study, we developed a new method that breaks down mixed meals reported in dietary recalls into their ingredients and groups these ingredients into a tree-based hierarchy. This hierarchical representation of ingredients was used to model the effect of inflammation in a large observational cohort and was compared with Dietary Inflammatory Index (DII) which uses nutrients (rather than intake of specific foods) as the current standard of assessing diet-induced inflammation. We observed that our tree-based model was better at predicting inflammation compared to the DII and could be more easily interpreted by showing specific foods (ingredients), rather than nutrients, that either increase or lower inflammation. These findings will help to provide better, more translatable, dietary guidance to prevent inflammation.

Technical Abstract: Background: Modeling the relationship between diet and inflammation is challenging with few existing methods to address this problem. Alternative representations of diet may help to improve predictions of health outcomes over traditional methods. Objective: To determine if hierarchical ingredient-level representations of diet improve predictive models of systemic inflammation in a large US cohort. Design: This population-based study used data on US adults (N=19,483) from the National Health and Nutrition Examination Survey (NHANES) 2001-2010 and 2015-2018 cycles. Results: The hierarchical representation of diet model (food tree) predicted class of systemic inflammation marker C-reactive protein (CRP) with higher accuracy (0.761) than the Dietary Inflammation Index (DII) (0.747). Individual dietary components revealed contributions towards increased inflammation including fruit punch, soda, and high-fat milk (OR: 0.001 – 0.005, P < 0.05), and foods associated with decreased inflammation such as herbal tea, coffee, brown rice, and pasta (OR: -0.08 – -0.001, P < 0.05). Conclusions: Specific ingredients, selected from a food tree, are stronger predictors of inflammation compared to the DII. Overall, our method provides better prediction and resolution to more precisely inform dietary guidance. Keywords: inflammation, dietary recall, hierarchical, etc.