Modelling the generation of flavour in a real food system

Control of aroma generation during the Maillard reaction presents great scientific and industrial interest. Although there have been many studies conducted in simplified model systems, the results are difficult to apply to complex food systems where the structure of the matrix and the presence of other components can have a significant impact on the reaction. In such complex systems, some degree of qualitative control may be possible, but quantitative and predictive changes can only be achieved through the development of mathematical models based on reaction kinetics. In the work described here, an aqueous extract of raw defatted homogenised meat was chosen as a simplified food matrix for studying the kinetics of the Maillard reaction. Beef liver extract was prepared and cooked for different time intervals from 5 to 240 minutes at 130C. Volatile analysis of the cooked samples was conducted using dynamic headspace extraction followed by GC-MS analysis. Free amino acid and sugar analyses were also performed. Volatiles identified included Strecker aldehydes, pyrazines and thiazoles. Multi-response kinetic modelling of the formation of 2and 3-methylbutanal was carried out based on a simplified version of the Strecker degradation pathway. Introduction The Maillard reaction is the reaction between reducing sugars and amino compounds which is responsible for much of the generation of flavour compounds in cooked foods. It consists of many parallel and consecutive pathways, which render it very complex and difficult to control. Currently, it is possible to manipulate certain volatiles by altering the selected precursors, but these changes are qualitative and are likely to have an impact on the wider volatile profile. Quantification of such changes is complicated as many factors are involved, and requires the development of mathematical models based on the kinetics of the Maillard reaction. Modelling is a process whereby a system of mathematical equations is generated in order to represent “reality” as accurately as possible. It is an iterative process (1) and comprises the following steps: generation of an initial hypothesis based on established reaction mechanisms, experimental design, modelling of data, model evaluation and criticism, and development of a new improved hypothesis on which to base further experiments. When the model is generated using data from several parts of the system simultaneously (reactants, intermediates, products), the process is called multi-response modelling. The aim of this work was to perform multiresponse kinetic modelling of the Maillard reaction in a simplified meat system in Expression of Multidisciplinary Flavour Science 285 order to predict the level of key flavour compounds for a given set of process conditions.