The use of metabolomics to predict cheese flavour development

Hajnajafi, A 2019, The use of metabolomics to predict cheese flavour development, Doctor of Philosophy (PhD), Science, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Hajnajafi.pdf Thesis application/pdf 3.07MB
Title The use of metabolomics to predict cheese flavour development
Author(s) Hajnajafi, A
Year 2019
Abstract This study characterized the flavour development of soft cheese during ripening, using Camembert as a model.  Gas chromatography-mass spectrometry (GC-MS), solid phase micro extraction (SPME) and Proton Nuclear Magnetic Resonance spectroscopy (1H NMR) were used to identify biomarkers of cheese flavour maturity. 1H NMR experiments were also performed to measure proteolysis and evaluate changes in amino acids levels during ripening. Near Infra-Red Spectroscopy (NIR) was used to develop new methods for predicting moisture, fat, protein and total solids which were then compared to traditional measuring techniques. The biochemical and physical data were then combined into a new chemometrics based model for predicting Camembert maturity.

The flavour profile of Camembert was determined at different temperatures (25, 35, 45 and 50 0C) by GC-MS/SPME. This study included the extraction, separation, and detection of volatiles at different time of ripening over a month. The optimized method for this work was determined to be using a DVB/CAR/PDMS fiber at 45ºC with a 30-minute extraction time. Based on the findings, the main volatile groups that contribute to flavour of Camembert cheese are esters, acids, ketones and alcohols.

A principal component analysis (PCA) was performed on each timepoint in the study to determine the most important volatile compounds at each stage of development. Principal components (PC) 1 and 2 explained 46.9% and 22% of the variability in the data respectively. The distribution of the scores on the first two PCs showed 3 separate groups, corresponding to different days of ripening. The data show that GC-MS-SPME-PCA is useful as potential tool for the control of Camembert cheese ripening because it is quick and non-invasive method of assessment that can provide a great deal of data.

A number of flavour compounds from different chemical groups were able to be identified and monitored through the maturation process with HS-SPME. This technique has been shown to provide a satisfactory evaluation of changes of the cheese during maturation and also to provide a good fingerprint of the compounds which are responsible for odour perception of cheeses by consumers.

The NIR models developed were based on real Camembert cheese samples at various stages of ripening. They were used to ascertain if NIR would be useful to measure fat, moisture, protein, and total solids. In order to have some data to compare with the NIR results standard measurements of each parameter were also performed on each sample. Statistical models were then built to compare the two data sets. 

A total of 6 training sets were developed for each of the four measurement types. The correlation coefficients between the two sets of data showed that NIR was just as good at measuring each parameter as standard techniques - and was also substantially faster. Based on the results outlined in this thesis, NIR spectroscopy could potentially improve dairy production through optimised laboratory efficiency and more efficient product quality control.

Results of the NMR analysis of the amino acids in cheese showed that significant differences in the levels of amino acids occur during ripening, resulting in markedly different metabolic profiles. Levels of leucine, valine, phenylalanine, methionine, asparagine, and glutamic acid increased as a function of ripening time, while threonine, proline, serine, aspartic acid, and histidine levels decreased. Together with the volatile analysis data, this work confirms the conversion of amino acids to aroma compounds by peptidases and proteinases in the cheese matrix during ripening. Hierarchical Cluster Analysis (HCA) showed that the proteolytic process was substantial in the first five days of ripening and then dropped off. The Biplot of the data clearly specified samples at different ripening stages were characterized by different amino acid profiles. The results proved that targeted 1H NMR metabolomics can provide a rapid approach to evaluate the fermentative process during cheese manufacturing, which may also provide a better understanding of how to control cheese quality.

A partial least squares (PLS) regression model was carried out to predict the age of Camembert using GC-MS-SPME, and ¹H NMR data to see if this form of analysis could replace the expensive and time-consuming sensory panels and cheese graders currently used in industry for this purpose. A PCA plot of physico-chemical data from both 1H NMR and GC-MS data sets showed a clear separation between samples during ripening. Using PRIMER it was shown that there was a significant relationship correlation between the physico-chemical and biochemical datasets.

The PLS coefficient plot obtained for physico-chemical variables showed that the most important variables that predicted the ripening time of Camembert cheeses were pH and water activity (aw). In the GC-MS/SPME data Hexanoic acid, Pentan-2-one, Heptan-2-one and Heptan-2-ol were the most important flavour compounds while the NMR showed that isoleucine and lysine were the most important amino acids. These data were combined for the final predictive model.

The prediction of the ripening time of the Camembert cheese by the PLS model was considered good with eight components selected by cross-validation, R2 (determination coefficient) = 0.967 and s (residual standard deviation) = 2.38 days. This new model that combines data and features gathered from multiple approaches and has clear potential to replace the expensive and time-consuming sensory panels currently used to predict and control cheese maturity and quality.

Overall, this study helps to increase our knowledge in soft cheese flavour profile development. The knowledge out of this project is likely to be of benefit to the dairy industry and dairy science in general.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Science
Subjects Separation Science
Food Chemistry and Molecular Gastronomy (excl. Wine)
Keyword(s) Flavour Development
Camembert cheese
Soft Cheese
Cheese Ripening
Version Filter Type
Access Statistics: 60 Abstract Views, 23 File Downloads  -  Detailed Statistics
Created: Tue, 24 Mar 2020, 09:58:27 EST by Keely Chapman
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us