Using blood metabolomics to identify dietary protein intake with Machine Learning methods

Publicerad

Typ

Examensarbete för masterexamen
Master's Thesis

Modellbyggare

Tidskriftstitel

ISSN

Volymtitel

Utgivare

Sammanfattning

This thesis examines how metabolomics data may be used to classify individuals based on the sources of protein in their diets. Developing accurate classification models that can distinguish between omnivores, vegans, vegetarians, and pescetarians is the aim of the study. Principal Component Analysis (PCA), Random Forest (RF), Support Vector Machines (SVM), and neural networks are used in this process as data analysis tools. The dataset, which was given by the Gothenburg University Department of Internal Medicine and Clinical Nutrition, included 120 healthy participants who followed various eating patterns. The subjects were chosen based on certain criteria, and blood samples and body composition were taken and examined. The dataset has been scaled and contains unidentified metabolites. The metabolic profile of the sample was shown using principal component analysis PCA). The overall PCA analysis revealed that there was substantial individual variation in the metabolomic profiles and that the food groups could not be effectively differentiated. The metabolic profiles of meat eaters and non-meat eaters might be used to distinguish them. Random Forest, SVM, and neural networks were the three machine learning techniques that were utilized for categorization. Neural Networks performed worse than Random Forest and SVM models in classifying each dietaryăgroup separately. Random Forest classified omnivores and non-omnivores with a high degree of accuracy. To measure the consumption of dairy, eggs, and meat, several scoring techniques were applied. The second method, which increased meat intake ratings by a factor of 1.5, produced the results with the highest degree of accuracy. This study sheds light on the metabolic effects of omnivorous diets and improves our understanding of the complex relationship between nutrition, metabolism, and health outcomes. It also highlights the potential of metabolomics and machine learning in predicting dietary patterns and categorizing people into different dietary categories.

Beskrivning

Ämne/nyckelord

metabolomics, machine learning, Principal Component Analysis, Random Forest, Support Vector Machines, Neural Networks

Citation

Arkitekt (konstruktör)

Geografisk plats

Byggnad (typ)

Byggår

Modelltyp

Skala

Teknik / material

Index

item.page.endorsement

item.page.review

item.page.supplemented

item.page.referenced