Analysis of the Hydrolytic Capacities of Aspergillus oryzae Proteases on Soybean Protein Using Artificial Neural Networks

Abstract

An artificial neural network (ANN) model was established to predict the hydrolytic capacities of Aspergillus oryzae proteases on soybean protein. The available training data were split in two subsets: training and testing data, which comprised 25 and six groups of proteases, respectively. These data served as the inputs of ANN to predict small peptide content, degree of hydrolysis and free amino nitrogen content. This network included three neurons in the single hidden layer with a low mean squared error. The predicted results were similar to the actual values (R 2 > 0.92) and were superior to those of multiple linear regression. Sensitivity analysis revealed that there is a correlation between protease and soy protein hydrolysates. It was also verified that protease and soy protein hydrolysates could serve as inputs and outputs in the ANN. Among the tested proteases, aminopeptidase showed the highest hydrolytic capacity for soybean protein with sensitivity analysis. PRACTICAL APPLICATIONSThe artificial neural network model is a powerful technique to predict the hydrolytic capacities of Aspergillus oryzae proteases for soybean protein. The results of this study could be used to test the amount of yielded hydrolysates of soybean protein under one combination protease, and also explained the mechanism underlying the protease-catalyzed hydrolysis of soybean.

Application of artificial neural network to predict the change of inosine monophosphate for lightly salted silver carp (hypophthalmichthys molitrix) during thermal treatment and storage

“…A back‐propagation algorithm (BP) was used to implement the supervised networks. This study used one hidden layer feed‐forward network structure with input and output layer, since one hidden layer has been found to be adequate in most cases (Li et al, ; Martinez et al, ). In order to better predict the percentage change of IMP, the number of neurons in the hidden layer was assigned through trial and error, as shown in Table .…”

Application of an artificial neural network model to predict the change of moisture during drying of sturgeon bone marrow

“…Only a single layer was used in this implementation because more layers may cause the local minimum problem (Rai et al, 2005;Li et al, 2016;Martinez et al, 2012). The optimal topology was chosen by determining the minimum error during testing.…”

Artificial neuronal networks (ANN) to model the hydrolysis of goat milk protein by subtilisin and trypsin

“…As for the protein hydrolysis, Abakarov et al (2011) satisfactorily modelled the kinetics of enzymatic hydrolysis of squid protein with subtilisin using the reaction time and the substrate concentration as input variables. Buciński et al (2008) and Li et al (2016) evaluated the variation of DH during the hydrolysis of bovine haemoglobin and pea proteins, respectively. Li et al (2006) developed a predictive model for the production of antioxidant peptides from fish proteins taking into account a number of input variables such as pH, temperature, hydrolysis time, muscle/water ratio and enzyme/substrate ratio.…”