Optimized parameters for the preparation of silk fibroin drug‐loaded microspheres based on the response surface method and a genetic algorithm–backpropagation neural network model

Abstract Using silk fibroin as the base material, the drug‐loaded microspheres are prepared by an emulsification method. In order to determine the drug‐loading and drug‐release performance parameters of the microspheres, the central composite design method is used to design and investigate the effects of the parameters of the microsphere preparation process, such as the oil–water ratio, stirring temperature, and stirring rate, on the microsphere particle size, drug‐loading rate, and drug release rate. The “overall desirability” is taken as a comprehensive evaluation index, and the response surface method (RSM) and genetic algorithm–backpropagation (GA–BP) neural network GA–BP model are used to predict and evaluate the parameters of the drug‐loaded microsphere preparation process. The root‐mean‐square error values obtained from the RSM and BP–GA model experiments are 0.000325 and 0.00022, respectively. The results show that the BP–GA model has better prediction accuracy and optimization ability than the RSM. The optimal microsphere preparation process conditions were determined to be as follows: a water–oil ratio of 10:1, at a temperature of 45°C with stirring at a speed of 400 rpm, the particle size of the microspheres is 1.392 μm, the drug‐loading rate is 3.218%, and the drug release rate is 51.991%. The results of this study indicate that this approach is an effective method for the optimization of the parameters of the drug‐loaded microsphere preparation process.