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

Back-propagation algorithm artificial neural networks (BP-ANN) were developed to predict the degradation of flavor precursor inosine monophosphate (IMP) trigged by thermal treatment at various temperatures (50, 60, 70, and 85°C) and durations in untreated (control) and 1.8% salted silver carp (T1) during refrigerated storage. IMP value of control and T1 decreased with heating time, with the fastest decline occurred at 50°C. When heated at 85°C for 10 min, IMP of control and T1 reached the maximum content after being stored at 4°C for 1 and 3 days. Salting treatment enhanced IMP content compared to untreated samples. The optimal BP-ANN models of control and T1 were able to predict the IMP change with correlation coefficients equaling 0.9715 and 0.9957 for testing phase. In conclusion, BP-ANN may be used as a potential tool for the prediction of IMP loss in silver carp during thermal treatments at day 0–3. Practical applications Silver carp is one of the most economically and nutritionally important freshwater fish species, with 4,967,739 tons in 2014 in the world. Nowadays, the ready-to-eat, lightly salted fish (∼2% NaCl) have become popular in the market. In the consumption of fish, Inosine monophosphate (IMP) contributes to the flavor and it can improve taste and flavor in cooked meat. In order to realize IMP variations of fish and define the maximum IMP concentration in samples during cooking and chilled storage, using models to predict the changes of IMP amount are of considerable interest. Therefore, this work was to establish model based on back-propagation artificial neural network to predict the IMP change in ready-to-eat salted silver carp for the fish industries or consumers in cold chain circulations.