Artificial neural networks to model the production of blood protein hydrolysates for plant fertilization

Abstract

BACKGROUND: Amino acid-based fertilisers increase the bioavailability of nitrogen in plants and help withstand stress conditions. Additionally, porcine blood protein hydrolysates are able to supply iron, which is involved in chlorophyll synthesis and improves the availability of nutrients in soil. A high degree of hydrolysis is desirable when producing a protein hydrolysate intended for fertilisation, since it assures a high supply of free amino acids. Given the complexity of enzyme reactions, empirical approaches such as artificial neural networks (ANNs) are preferred for modelisation. RESULTS: Porcine blood meal was hydrolysed for 3 h with subtilisin. The time evolution of the degree of hydrolysis was successfully modelled by means of a feedforward ANN comprising 10 neurons in the hidden layer and trained by the Levenberg-Marquardt algorithm. The ANN model described adequately the influence of pH, temperature, enzyme concentration and reaction time upon the degree of hydrolysis, and was used to estimate the optimal operation conditions (pH 6.67, 56.9 °C, enzyme to substrate ratio of 10 g kg-1 and 3 h of reaction) leading to the maximum degree of hydrolysis (35.12%). CONCLUSIONS: ANN modelling was a useful tool to model enzymatic reactions and was successfully employed to optimise the degree of hydrolysis.