基于谱效关联及AHP-EWM综合赋权的Box-Behnken响应面法结合GA-BP神经网络多指标优化白芍甘草提取工艺

Box-Behnken Response Surface Method Based on Spectrum-Effect Correlation and AHP-EWM Comprehensive Weighting Combined with GA-BP Neural Network for Optimizing the Extraction Process of Radix Paeoniae Alba and Glycyrrhiza

关键词:  白芍甘草  谱效关联  Box-Behnken响应面法  AHP-EWM综合赋权  GA-BP神经网络

Key words:  Radix Paeoniae Alba and Glycyrrhiza  spectrum-effect correlation  Box-Behnken response surface method  AHP-EWM comprehensive weighting  GA-BP neural network

摘要:

目的采用谱效关联及AHP-EWM综合赋权的Box-Behnken响应面法结合GA-BP神经网络建立白芍甘草药对的综合评价指标,确定白芍甘草药对的最佳提取工艺。方法 采用Box-Behnken响应面法考察料液比、超声时间、乙醇浓度对白芍甘草提取工艺的影响,测定白芍甘草提取物干膏得率及其DPPH·清除率,建立其HPLC指纹图谱,选取没食子酸、芍药内酯苷、芍药苷、甘草苷、异甘草苷、甘草素、异甘草素、甘草酸等8种已知入血成分作为白芍甘草提取物的主要质量标志物。采用灰关联度法分析指纹图谱与DPPH·清除率间的谱效关系,计算关联度,获取关联度校正后的总峰面积和8种指标成分峰面积。采用AHP-EWM法对评价指标进行综合赋权并计算综合评价指标,获取Box-Behnken响应面优化的白芍甘草最佳提取工艺,并与建立的GA-BP神经网络模型所预测的最佳提取工艺比较、验证。结果 料液比、超声时间、乙醇浓度因素下的关联度变化范围分别为0.7174～0.8654、0.6679～0.8721、0.6436～0.8511； Box-Behnken 响应面法和GA-BP神经网络优化的综合评价指标分别为0.9843(RSD=3.2%)、0.9255(RSD=2.2%),经对比及验证,确定白芍甘草的最佳提取工艺为料液比1∶40(g·mL-1),超声时间45 min,乙醇浓度75.29%。结论 基于谱效关联及AHP-EWM综合赋权的Box-Behnken响应面法结合GA-BP神经网络多指标确定了白芍甘草的最佳提取工艺,也为其他中药提取工艺的优化提供了新思路。

Abstract:

ABSTRACT: OBJECTIVE To optimize the extraction process of Radix Paeoniae Alba and Glycyrrhiza by the Box-Behnken response surface method based on spectrum-effect correlation and AHP-EWM comprehensive weighting combined with GA-BP neural network. METHODS During the experiment, the effects of material-liquid ratio, ultrasound time and ethanol concentration on the extraction process of Radix Paeoniae Alba and Glycyrrhiza were investigated by using Box-Behnken response surface methodology. The dry extract yield of Radix Paeoniae Alba and Glycyrrhiza was calculated, its DPPH· clearance rate was measured and its HPLC fingerprint was established. Eight known blood components, including gallic acid, paeonilactone glucoside, paeoniflorin, glycyrrhizin, isoglycyrrhizin, glycyrrhizin, isoglycyrrhizin and glycyrrhetinic acid were selected as the main quality markers of Radix Paeoniae Alba and Glycyrrhiza extracts. The spectrum-effect relationship between fingerprint and DPPH· clearance rate was analyzed using grey correlation method, and the correlation degree was calculated. Peak area in fingerprint was proofread using correlation degree as weight coefficient, and total peak area and the peak area of 8 indicator components were obtained after proofreading. The comprehensive evaluation indicators were calculated after comprehensively weighting the evaluation indicators using AHP-EWM method to obtain the best extraction process of Radix Paeoniae Alba and Glycyrrhiza with Box-Behnken response surface optimization and GA-BP neural network model was established to predict and further validate the extraction process of Radix Paeoniae Alba and Glycyrrhiza.. RESULTS The range of correlation degree changes under the factors of material-liquid ratio, ultrasound time and ethanol concentration was 0.7174-0.8654, 0.6679-0.8721 and 0.6436-0.8511, respectively. The comprehensive evaluation indicators of Box-Behnken response surface method and GA-BP neural network optimization were 0.9843 (RSD=3.2%) and 0.9255 (RSD=2.2%), respectively. After comparison and validation, the optimal extraction process for Radix Paeoniae Alba and Glycyrrhiza was determined as follows: The solid-liquid ratio was 1∶40 （g·mL-1）, the ultrasound time was 45 minutes and the ethanol concentration was 75.29%.CONCLUSION In summary, Box-Behnken response surface method based on spectrum-effect correlation and AHP-EWM comprehensive weighting combined with GA-BP neural network has optimized the extraction process of Radix Paeoniae Alba and Glycyrrhiza, which also provides new ideas for optimizing other traditional Chinese medicine extraction processes.