Artificial neuronal network modeling of the enzymatic hydrolysis of horse mackerel protein using protease mixtures

作者：Morales Medina R^*; Perez Galvez R; Guadix A; Guadix E M

来源：Biochemical Engineering Journal, 2016, 105: 364-370.

DOI：10.1016/j.bej.2015.10.009

• Bioreactions
• Enzyme bioreactors
• Modeling
• Optimization
• Proteolysis
• Artificial neural networks

In enzymatic reactions, the modeling of the degree of hydrolysis (DH) is desirable given its influence on many functional and biological activities. Empirical approaches are preferred to phenomenological ones to model the complexity of enzymatic reactions. Artificial Neuronal Networks (ANN) are able to process large sets of data where no linear relationships between them are expected. In this work, a feedforward ANN, comprising 10 neurons in the hidden layer, was successfully employed to model the DH as a function of the initial concentration of horse mackerel protein (tested at 2.5 g/L, 5 g/L and 7.5 g/L), the reaction temperature (40 degrees C, 47.5 degrees C and 55 degrees C), the time of reaction (up to 4h) and the percentage of subtilisin in the enzyme mixture (0%, 25%, 50%, 75% and 100%). The resulting ANN model was optimized by an evolutionary algorithm, obtaining a maximum (DH 17.1%) at 2.54 g/L, 40 degrees C, 4h and an enzyme mixture comprising 38.3% of subtilisin and the rest of trypsin. The combination of trypsin and subtilisin led to higher DH than the sole use of subtilisin (DH 15.5% at 52 degrees C). Furthermore, the former optimum was attained at lower reaction temperature, which reduces both the operational costs and the nutritional losses.

• 出版日期2016-1-15