Flavor enhancement of Perinereis aibuhitensis Maillard reaction liquids based on the RSM-BP-GA model

The application of Perinereis aibuhitensis is limited by the poor flavor of its hydrolysate. Thus, we aimed to optimize the preparation conditions of P. aibuhitensis Maillard reaction solution (PMS) by using a response surface methodology (RSM) combined with a back-propagation neural network coupled with a genetic algorithm (BP-GA). The optimal conditions were as follows: ribose:glucose, 1:1; reducing sugar addition, 3.21%; pH, 6.5; reaction temperature, 121 °C; and reaction time, 64 min. The volatile flavor components and free amino acid content in PMS (prepared under the optimal conditions) and P. aibuhitensis enzymatic hydrolysis solution were determined. PMS had good flavor, reduced aromatic hydrocarbon and ether contents, and increased aldehyde. The proportions of bitter and sweet amino acids decreased and increased, respectively, which significantly improved taste. This study demonstrates the effectiveness of machine learning-assisted optimization for flavor enhancement and provides a viable strategy for developing high-value seafood seasonings from underutilized marine resources. • Preparation conditions for PMS were optimized using RSM combined with BP-GA. • PMS prepared under optimized conditions had reduced undesirable flavor substances. • PMS prepared under optimized conditions had more pleasant aroma volatile components. • PMS prepared under optimized conditions had significantly improved flavor. • Machine learning algorithms can improve the quality of food products.