Optimization of the formulation of corn-based gluten-free bread with egg white protein and the microbial transglutaminase enzyme and parameters estimation of this process with artificial neuron network

In this research, optimization of the formulation of corn-based gluten-free bread with egg white protein and the microbial transglutaminase enzyme was conducted with the purpose of minimizing baking loss and hardness, and maximizing the specific volume, bread height, porosity and color index L* by the response surface method and finally, estimating these responses with the help of the neural network. The effects of the two concentrations of microbial transglutaminase enzyme in the range of 0 to 1.5% and egg white protein in the range of 0 to 6% with the help of the central composite design on specific volume parameters, baking loss, porosity, bread height, hardness and color indices L*, a* and b* of bread crust were investigated. Investigating the results indicated that the optimum conditions for the production of gluten-free bread based on maize is created when the enzyme and egg white protein concentrations are 0.65% and 6%, respectively. By increasing the enzyme concentration, baking loss, porosity and a* index increased, but the b* index decreased. The results also indicated that by increasing the enzyme concentration in the formulation of samples, we firstly observed increase and then decrease in bread height and specific volume. Increasing the protein concentration of egg white in maize-based gluten-free bread formulation resulted in an increase in L* and a* indexes, and a decrease in baking loss and b* index. The results of artificial neural network modeling indicated that a network with a hidden layer containing 9 neurons, with an 8-9-2 layout (a network with 2 inputs, 9 nodes (neurons) in the hidden layer and 8 outputs), had the best result in predicting the mentioned outputs. This network indicated the highest accuracy among the mentioned topologies with a correlation coefficient of 1.00 and mean square error of 0.0011.