Integrating NMR-Based Metabolic Phenotyping and AI Approaches for Dietary Biomarkers Discovery and Nutritional Assessment

Andy Zhu

doi:10.60867/00000063

Non-communicable diseases (NCDs), such as cardiovascular disease, diabetes, and cancers, impose a major global health burden, with diet being a critical determinant of risk. While dietary patterns such as the Mediterranean and Dietary Approaches to Stop Hypertension (DASH) diets are associated with reduced disease risk, the mechanistic understanding of how specific foods influence health outcomes remains limited, yet this knowledge is essential for advancing precision nutrition. This thesis integrates molecular phenotyping, dietary intervention studies, and computational text mining approaches to improve nutritional assessment and advance dietary biomarkers discovery. Nuclear magnetic resonance (NMR)-based lipoprotein profiling demonstrated strong agreement with standard clinical assay, confirming the robustness of NMR while highlighting areas where methodological harmonisation would be beneficial (Chapter 3). Dietary intervention studies explored the potential for identifying NMR-based biomarkers of food intake (Chapter 5). Both established and novel urinary biomarkers for commonly consumed fruits and vegetables were identified, providing objective tools for dietary exposure assessment and enabling more precise evaluation of diet-health interactions in epidemiological study. Additionally, these interventions facilitated investigation into how inulin influences gastrointestinal transit and fermentation (Chapter 4), underscoring the importance of food molecular structure and microbial functionality in shaping metabolic responses. Complementing experimental work, artificial intelligence (AI) methods were applied to extract and classify context-aware food-disease relationships from biomedical literature using chainof- thought prompting strategies (Chapter 6). Large language models (LLMs) achieved high accuracy in distinguishing beneficial from harmful associations, demonstrating the potential of AI-enable evidence synthesis to scale nutritional research and generate testable hypotheses. Future research should focus on validating identified dietary biomarkers in diverse populations, expanding AI-driven approaches to map food-health relationships across a broader range of diseases, and integrating biomarkers data with metabolic pathway analysis. These efforts will deepen mechanistic understanding and support the development of more precise and personalised nutrition strategies. Together, these contributions establish a multidimensional framework that integrates NMRbased molecular phenotyping, dietary biomarker discovery, and AI-assisted literature mining to advance nutrition science. This work provides practical, scalable approaches for improving dietary assessment and enhancing understanding of diet-health interactions, ultimately supporting the development of targeted public health recommendations and personalised nutrition strategies.

Integrating NMR-Based Metabolic Phenotyping and AI Approaches for Dietary Biomarkers Discovery and Nutritional Assessment

Files and links (1)

Abstract

Details

UN Sustainable Development Goals (SDGs)

Metrics