BioModels Database logo

BioModels Database

spacer

Mellor et al., (2010). Reduction of off-flavor generation in soybean homogenates: a mathematical model.

December 2012, model of the month by Christine Seeliger
Original model: BIOMD0000000415


Soybeans are legumes known as food crops in Asia for thousands of years. Despite industrial production and use as animal feed they became known as a good substitute for animal products in the human diet outside of Asia in the last century [1]. The United States, Brazil and Argentina are the main producers of soy, producing 81% of the worlds soybean harvest on about 6% of the overall arable land. Soybeans are a source of complete protein with about 38% of a soybean seed being proteins. This renders them a frequent component of meat and dairy analogues i.e. in the form of TVP (Textured Vegetable Protein).

However, soybeans need to be processed to create a variety of desired soy products. The beans are crushed to produce a homogenate that can be processed further. Certain enzymes and their substrates that were separated in the cells beforehand get together and produce substances that result in so-called undesirable off-flavours [2]. Frequently, the lipoxygenase (LOX) pathway [3] is responsible for the production of these off-flavours. Soybeans have a high content in linoleic acid, a suitable substrate for lipoxygenases to formn-hexanal. Although n-hexanal is used as food flavour, it is desirable to reduce and control its production during soybean processing, to create a product that does not have too much variation in flavour.

Soybeans contain three different LOX isozymes. The two intermediates 13HOD-S(Z,E), and 13HOD-R(Z,E) are formed in different proportions at different rates by these three isozymes (named L1-3). They are further processed by a hydroperoxide lyase (HPL) to form n-hexanal (see Figure 1). Different soybean strains have been engineered that contain different combinations of the three LOX enzymes. Their production of n-hexanal has been experimentally analysed in previous studies.

Figure 2

Figure 2 Comparison of model n-hexanal concentration for (A) various simulated null beans and (B) experimentally obtained results (Matoba and others 1985a), in which no standard errors were given. Figure taken from [4].

Comparison between the n-hexanal time courses between simulation [Figure 2a] and experiment [Figure 2b] indicate that the simulation captures the order of maximum n-hexanal production between different bean strains well but fails to point out the much higher n-hexanal concentrations produced in experiments by the L2 only beans [Figure 3]. However, the experiments were done at different pH and the pH has a high influence on the analysed reactions. Unfortunately, no experiments were done so far to test the concentrations of the different intermediates and products in a systematic manner to address the influence of pH changes. The authors point out that an additional pathway degrades n-hexanal to n-hexanol at higher pH, that is not accounted for in the model but might be useful in controlling flavour in soyproducts as n-hexanol has a much weaker flavour than n-hexanal. The authors tested a range of different Km parameters for the enzymes. Reducing L3 Km and increasing L2 Km seem to have the biggest effect on n-hexanal production [Figure 4].

Overall the results suggest that decreasing L2 concentrations and/or increasing L3 could provide measures to keep the n-hexanal production in soybean homogenates down. However, the model so far fails to accurately capture the behaviour of L1 beans which might offer interesting solutions as well. Adapting the rates of different enzymes helps to obtain an overall better fit of the model to experimental data, but does not fix the behaviour of L1 which is supposed to produce much less n-hexanal in real soybean homogenates than in the simulation. The authors also point out that including the dependency on pH might be a vital step to improve the model and further analyse the different contributions of the single enzymes. This will help to identify the ideal conditions with regard to n-hexanal production in the homogenates.

Models like this help to dissect the different components of signalling pathways and address the effect that each single component has on the overall outcome. In this case the production of an unwanted metabolite. This study illustrates the application and provides insides in the possible value of theoretical approaches to optimise production methods in the food industry.

Figure 1

Figure 1 Metabolic LOX pathway in soybeans. Figure taken from [1].

Mellor et al. (2011) [4, BIOMD0000000415] authors used this data to construct and validate an ODE based mathematical model to further study n-hexanal production and devise new methods to reduce and control it. The metabolic pathways implemented in Copasi are shown in Figure 1. Their model sufficiently captures different aspects of the experimental observations. The highest concentration of n-hexanal is produced by L2 only beans, closely followed by those beans that contain L2 in different combinations with the other enzymes: L12, L23 and wildtype beans. L3 beans produce by far the lowest amount. In most cases, the total n-hexanal concentration is close to experimental results, however L1 beans show the biggest discrepancy, with simulated concentrations being far too high. In experiments, L1 and L3 beans produce similar low amounts of n-hexanal (although L1 LOX activity is 3 times as high as for L3 and L1 products contain favourable 13HOD-S(Z,E) isoform in much higher proportion than L3 products). The authors speculate that the reason might be a missing feedback in the model or the chosen kinetics are not accurate.

Figure 3

Figure 3 Individual comparison between model predictions and experimental results for L2 beans. Figure taken from [4].

Figure 4

Figure 4 Simulated peak n-hexanal concentration against varying Km parameter values for L-1, L-2, and L-3 isoenzymes in "wild type" beans. Peak n-hexanal is given relative to the value at default Km (0.49 mM). Figure taken from [4].

Bibliographic references

  1. Hartman et al. Crops that feed the World 2. Soybean—worldwide production, use, and constraints caused by pathogens and pests. Food Security. 2011; 3(1):5-17.
  2. Lei and Boatright. Compounds contributing to the odor of aqueous slurries of soy protein concentrate. Journal of Food Science. 2001; 66(9):1306-1310.
  3. Feussner and Wasternack. The lipoxygenase pathway. Annual Review of Plant Biology. 2002; 53:275-297.
  4. Mellor et al. Reduction of off-flavor generation in soybean homogenates: a mathematical model. Journal of Food Science. 2010; 75(7):R131-138
spacer
spacer