Optimal extraction bioactive components of tetramethylpyrazine in Chinese herbal medicine jointly using back propagation neural network and genetic algorithm in R language.

A combinational approach of back propagation neural network (BPNN) and genetic algorithm (GA) was proposed in the present study to optimize the extraction technology of tetramethylpyrazine (TMP) in Ligusticum wallichii Franchat. Based on the single factor test, the orthogonal experiment design method of four factors and three levels was adopted, and the concentration of TMP was measured by high performance liquid chromatography (HPLC). Subsequently, BPNN model was trained for a predictive computational model of the performance indices via experimental data, and GA was exploited to find the optimization con ditions for extraction technology of TMP. Meanwhile, both the model and algorithm were implemented in R language. Ethanol concentration of 80%, extraction time of 1.5h, extraction temperature of 55 and liquid-solid ratio of 8:1 were derived as optimal conditions with a maximum content of TMP of 2.04 mg/g, which was confirmed with the relative error 2.63% through the validation of the experiments. This mathematical model could be used to analyze and predict the extraction technology of TMP in Ligusticum wallichii Franchat and provide a new reference for screening optimization of Chinese medicine effective parts and components.