基于人工神经网络和遗传算法优化灵芝液态发酵培养基  

Optimization of Ganoderma lucidum Liquid Fermentation Medium Based on Artificial Neural Network and Genetic Algorithm

作　　者：郭嘉[1,2] 刘艳芳 唐传红[2] 冯杰[2] 张劲松[2] GUO Jia;LIU Yanfang;TANG Chuanhong;FENG Jie;ZHANG Jingsong(University of Shanghai for Science and Technology,School of Health Science and Engineering,Shanghai 200093,China;Institute of Edible Fungi,Shanghai Academy of Agricultural Sciences,Key Laboratory of Edible Fungi Resources and Utilization(South),Ministry of Agriculture and Rural Affairs,P.R.China,National Engineering Research Center of Edible Fungi,Shanghai 201403,China)

机构地区：[1]上海理工大学健康科学与工程学院,上海200093 [2]上海市农业科学院食用菌研究所,国家食用菌工程技术研究中心,农业农村部南方食用菌资源利用重点实验室,上海201403

出　　处：《食用菌学报》2023年第3期81-89,共9页Acta Edulis Fungi

基　　金：上海市现代农业产业技术体系[沪农科产字(2022)第9号];上海市农业科学院卓越团队建设计划[沪农科卓(2022)003];上海市农业科学院攀高计划(2021-2024)。

摘　　要：以灵芝(Ganoderma lucidum)胞内相对分子质量较高的多糖组分的峰面积为指标,采用单因素实验筛选液态发酵培养基的氮源及其浓度,按照Box-Behnken设计29个实验组,并进行发酵培养实验,建立数据集;将数据集随机分为训练集(24组)和测试集(5组),训练集用于构建BP神经网络模型,测试集用于评估模型,采用遗传算法对BP神经网络的拟合结果进行寻优。结果表明:模型拟合度可达0.9848;最佳培养基组成为29.0 g·L^(-1)无水葡萄糖、3.7 g·L^(-1)酵母自溶粉、0.3 g·L^(-1)MgSO_(4)·7H_(2)O,采用优化后的发酵培养基培养比未优化的基础培养基培养获得的相对分子质量较高的多糖峰面积提高40.8%。Using peak area of a high relative molecular weight intracellular Ganoderma lucidum polysaccharide fraction(26–36 min in HPSEC)as the index,nitrogen source and concentration in the liquid fermentation medium were screened by single-factor experiments.Then 29 fermentation groups were conducted according to a BoxBehnken design to establish a data set,which was then randomly divided into a training set(24 groups)and a testing set(5 groups).The training set was used to construct a BP neural network model and the testing set was used to evaluate the model.The BP neural network fitting results were further iteratively optimized by a genetic algorithm.The results showed that the model fit reached 0.9848,and the optimal medium composition was 29.0 g·L~(-1)anhydrous glucose,3.7 g·L~(-1)autolyzed yeast extract,and 0.3 g·L~(-1)MgSO_4·7H_2O.Using the optimal medium,the peak area of the high molecular weight intracellular polysaccharide fraction was 1.20×10~7 mV·s,which was40.8%greater than that before optimization.

关 键 词：灵芝 液态发酵 胞内多糖 相对分子质量 人工神经网络 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程] TS219[自动化与计算机技术—控制科学与工程]