张 峰，吴柳娟，李 躺，王 灿，胡益波，丁学知，夏立秋

(湖南师范大学生命科学学院，微生物分子生物学湖南省重点实验室，省部共建淡水鱼类发育生物学国家重点实验室，中国 长沙 410081)

应用遗传算法和神经网络优化多杀菌素发酵培养基

张 峰，吴柳娟，李 躺，王 灿，胡益波，丁学知，夏立秋

(湖南师范大学生命科学学院，微生物分子生物学湖南省重点实验室，省部共建淡水鱼类发育生物学国家重点实验室，中国 长沙 410081)

通过优化刺糖多孢菌发酵合成多杀菌素培养基成分，改善培养条件，从而提高多杀菌素产量.在单因素以及Plackett-Burman试验设计的基础上，采用Box-Behnken试验设计方法对发酵培养基组分中的玉米浆、可溶性淀粉、丙酸钠进行研究，运用遗传算法优化的BP神经网络建立多杀菌素产量与培养基组分浓度之间的预测模型，采用循环算法对此模型进行寻优，得到三种组分的最佳配比为：玉米浆7 g/L、可溶性淀粉16 g/L、丙酸钠2 g/L，多杀菌素产量达到(550.22±3.84)mg/L，采用上述方法优化后的培养基使得多杀菌素产量比原始培养基产量(225 mg/L)提高145%.本研究结果可为培养基优化提供一种有效的建模方法.

刺糖多孢菌；多杀菌素；发酵培养基优化；神经网络；遗传算法

AbstractIn this work, we report an approach to improve spinosad production by optimizing the fermentation medium components. Corn steep liquor, soluble starch and sodium propionate in the fermentation medium were investigated by Box-Behnken design (BBD), which was based on single factor and the Plackett-Burman design. Moreover, a prediction model of the spinosad yield as a function of the medium component concentration has been established by using the artificial neural network (ANN) optimized by genetic algorithm (GA). Using ANN optimized by GA as the objective function, we employed the circulatory algorithm to optimize the medium components and the optimal ratio of the three components as follows: with corn steep liquor 7 g / L, soluble starch 16 g/L, and sodium propionate 2 g/L, the final yield of spinosad reached (550.22±3.84) mg/L, which was 145% higher than the original fermentatiom medium(225 mg/L)obtained when cultured on the optimized medium. Our results from this study can provide an effective modeling method for medium optimization.

Keywordssaccharopolysporaspinosa; spinosad; optimization of fermentation medium; neural network; genetic algorithm

多杀菌素又名刺糖菌素，是由土壤放线菌刺糖多孢菌 (Saccharopolysporaspinosa)经有氧发酵产生的具有大环内酯结构的次级代谢产物[1].因其具有广谱杀灭害虫作用，对人、非靶标动物和环境安全，可生物降解等优点[2]，被国际上认为是一种具有发展前景的绿……