Epsilon Archive for Student Projects

Abstract

Rye products are known to more effective decrease different aspects of subjective appetite compared to low fiber wheat bread. Nutritional composition and structural properties, such as dietary fiber components and grain integrity are thought to play roles in the satiating mechanism behind this appetite reduction. The aim of this master project was to investigate how a whole grain rye porridge breakfast with addition of plant protein and easily fermented fiber is reflected in the short term plasma metabolome and if there are differences in the metabolome that may explain differences in perceived appetite in healthy people.
The study design was a single blinded randomized cross-over with six treatments tested on healthy men and women (n = 21). The tested meals were iso-caloric serv-ings of plain rye porridge made of 40 g rye (R), 55 g rye (R55), or R with three levels of easily fermented dietary fiber and plant protein; 9 g fiber/ 3 g protein (RHF), 6 g fiber/6 g protein (RPF), and 3 g fiber/9 g protein (RHP). White wheat bread (WWB) served as control. Plasma metabolome was studied at 8 time points over an 8 hour period containing a standardized lunch and a dinner with unlimited supply. Perceived appetite was assessed with visual analogue (VAS) scale every 30 minutes.
Untargeted NMR-based metabolomics was applied on ultra-filtrated, protein free, plasma samples. To facilitate the primary aim the automated NMR peak picking and alignment treatment methods of “speaq” (Spectral Alignment and Quantifica-tion) package in the software programming language R were developed and evalu-ated. Prior to “speaq” processing NMR data were subjected to traditional manual phase and baseline correction in TopSpinTM 2.0 (Bruker Corporation, USA). The processed NMR dataset was analyzed with principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) to seek correlations between the metabolome and the rye breakfasts.
RHF, RPF, RHP and R55 were found to reduce the feeling of hunger over 8 hours (Least Square Means) compared to WWB. The final data treatment workflow resulted in > 1200 detected features. Of those, a vast quantity was noise and many peaks were missing, but still a lot of useful information was generated in a fraction of the time required for current manual peak picking methods. PCA and PLS-DA models on entire dataset or subsets were distinguished by time points, individuals, or baseline differences between occasions, rather than by treatment. The current methods could not distinguish between metabolomes of any of the six treatments.
In this project several important steps towards an automated processing of NMR data were made. Further research on improvement of the “speaq” workflow in gen-eral and peak picking in particular are suggested, as well as accurate assessment in the statistical models of the large variation not related to the diet intervention.