Peptide fingerprinting and predictive modelling of fermented milk : a thesis presented in partial fulfilment of the requirements for the degree of Doctor of Philosophy in Food Technology at Massey University, Palmerston North Campus, New Zealand
Loading...
Date
2021
DOI
Open Access Location
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Massey University
Rights
The Author
Abstract
Fermented milk products are valued by consumers and the food industry for their nutritional properties, pleasant taste, and texture. Consumer demands and expectations for such products are constantly changing. Understanding how consumers perceive the sensory characteristics of food and the relationship these characteristics have with the chemical components of food can provide insight that can enable food researchers and manufacturers to develop food products that are tailored to provide enhanced sensory qualities. Establishing techniques that allow for in-silico prediction or correlation of sensory qualities can enable a more rapid approach that would aim to enable researchers to meet the demands of consumers.
This research firstly explored mass spectrometric techniques for the rapid fingerprinting of milk and fermented milk products, using Matrix-Assisted Laser Desorption Ionisation - Time-of-Flight Mass Spectrometry (MALDI-TOF MS) and Rapid Evaporative Ionisation Mass Spectrometry (REIMS), two technologies that require minimal sample preparation and can rapidly generate a fingerprint of a food’s chemical components. Peptide fingerprints obtained by MALDI-TOF MS and analysed by principal component analysis were effective at discriminating the two fermented milk and milk samples. Supervised discrimination of low molecular weight fingerprints obtained via REIMS and MALDI-TOF MS proved less effective but demonstrated some potential and could be used alongside other analyses in future studies. These techniques were explored with a view to establishing a technique that could provide rapid insights into a food’s chemical composition, and which could also effectively discriminate the chemical components of the product. Such techniques could be used for rapid screening of products and can provide insight into the chemical components that are driving the variation in different products, which may be reflective of the differences in sensory characteristics.
Next, peptide fingerprinting and predictive modelling were investigated in milk fermented with various bacterial combinations, including probiotic cultures. Fingerprinting was performed on samples collected at each hour of fermentation. Predictive modelling techniques, using both regression and classification approaches, were trialled in order to predict the change in signal intensity throughout fermentation. This aimed to understand if peptides could be predicted throughout fermentation, with a view to enable the targeted prediction of desirable peptides, or other relevant components, which may impart favourable sensory qualities in the final product. Regression techniques were somewhat effective for predicting the signal intensity of individual m/z ions throughout fermentation. Most of the ions did not follow a linear relationship, and, as such, a multiple linear regression model was unable to model most of the ions. Using a generalised additive model, a non-linear approach, improved the performance in most cases and could predict the signal intensity of individual ions throughout fermentation. However, the model was unable to correctly predict all cases. Classification techniques were effective for predicting the general direction of the signal intensity between start and end fermentation times. Five classification techniques were trialled, with each model providing accurate predictions for the increase or decrease of signal intensity between early and late fermentation times.
Lastly, consumer panellists were recruited to evaluate the change in important sensory characteristics throughout the fermentation of milk prepared using two different starter cultures. This aimed to understand if consumer responses to such products could be correlated with instrumental analysis, in order to predict the consumer responses from instrumental data. Consumers perceived significant differences in bitterness and flavour intensity between fermented milk samples at different fermentation time points. There were significant correlations between peptide fingerprints and the consumer rankings for the sensory attributes in each fermented milk product. XGBoost regression could predict consumer responses with reasonable accuracy.
This thesis explored the fermentation of milk using specific bacteria and fermentation processes. To validate this work, further products could be explored, in addition to different processing parameters. Furthermore, a more in-depth analysis of the chemical components of the products could be investigated and analysed with additional sensory evaluation to further explore and confirm the findings.
Description
Keywords
Fermented milk, Composition, Peptides, Mass spectrometry, Fermentation, Mathematical models, Dairy products, Sensory evaluation