引用本文:	张静,王立军,羌玺,黄丹琳,解修俊,王旭雷,马增岭,王广策.基于响应面和人工神经网络的紫菜中类菌孢素氨基酸提取工艺优化[J].海洋科学,2025,49(12):115-127.

【打印本页】【HTML】【下载PDF全文】【查看/发表评论】【下载PDF阅读器】【关闭】

过刊浏览高级检索

本文已被：浏览 165次下载 43次	码上扫一扫！
字体:加大+\|默认\|缩小-
基于响应面和人工神经网络的紫菜中类菌孢素氨基酸提取工艺优化
张静^1,2, 王立军², 羌玺², 黄丹琳³, 解修俊², 王旭雷², 马增岭¹, 王广策^2,4
1.温州大学生命与环境科学学院, 浙江温州 325035;2.中国科学院海洋研究所实验海洋生物学实验室, 山东青岛 266000;3.哈尔滨工程大学材料科学与化学工程学院, 黑龙江哈尔滨 150001;4.水产品种创制与高效养殖重点实验室(中国科学院), 山东青岛 266000

摘要:

类菌孢素氨基酸(mycosporine-like amino acids, MAAs)是一类具有防紫外、抗氧化和抗炎等活性的化合物,在化妆品、食品及医药等领域具有重要用途。海洋生物,尤其是海洋藻类,大多含有类菌孢素氨基酸,其中红藻中的条斑紫菜(Pyropia yezoensis)是类菌孢素氨基酸含量较高的种类之一。本文基于响应面和人工神经网络分析方法,对条斑紫菜中类菌孢素氨基酸提取工艺进行优化:以类菌孢素氨基酸提取率为指标,在单因素实验考察基础上设计构建响应面模型和人工神经网络模型,并对2种模型方法进行比较分析和验证。研究表明, 2种模型分析结果均显示料液比和提取时间对提取率的影响最为显著;响应面模型具有较好的预测能力与可靠性,优化条件下实验值(0.468%)与预测值(0.474%)高度吻合;人工神经网络模型在训练集与验证集上表现出较高的拟合优度(R²>0.93),但其在独立测试集上的预测误差较大,泛化能力不及响应面模型;最后确定最佳提取条件为:温度42℃、时间1.51 h、料液比1∶25(g/mL)、提取2次、甲醇浓度20%,在此条件下提取率达0.468%。该研究将为条斑紫菜中类菌孢素氨基酸的规模化提取与利用提供重要参考。

关键词: 类菌孢素氨基酸条斑紫菜提取工艺响应面人工神经网络

DOI：10.11759/hykx20260202001

分类号:P745

基金项目:山东省重点研发计划(重大科技创新工程)项目(2025CXGC010618,2025LZGC037);国家自然科学基金(42276146,41876124);浙江省自然科学基金会(LZ21C030001);农业农村部(CARS-50);青岛市关键技术攻关项目(25-1-1-gjgg-54-hy);山东省“泰山学者”工程专项经费资助项目(tspd20210316)

Optimization of the extraction process of mycosporine-like amino acids from Pyropia yezoensis using response surface methodology and artificial neural networks

ZHANG Jing^1,2, WANG Lijun², QIANG Xi², HUANG Danlin³, XIE Xiujun², WANG Xulei², MA Zengling¹, WANG Guangce^2,4

1.School of Life and Environmental Sciences, Wenzhou University, Wenzhou 325035, China;2.Laboratory of Experimental Marine Biology, Institute of Oceanology, Chinese Academy of Sciences, Qingdao 266000, China;3.College of Materials Science and Chemical Engineering, Harbin Engineering University, Harbin 150001, China;4.Key Laboratory of Breeding Biotechnology and Sustainable Aquaculture (CAS), Institute of Oceanology, Chinese Academy of Sciences, Qingdao 266000, China

Abstract:

Mycosporine-like amino acids（MAAs） are a class of organic compounds that have been demonstrated to exhibit UV-protective, antioxidant, and anti-inflammatory properties. They have remarkable applications in the cosmetics, food, and pharmaceutical industries. A multitude of seaweed species have been found to contain MAAs, with red alga Pyropia yezoensis（P. yezoensis） being a particularly noteworthy example because of its remarkably high MAA content. The present study employed response surface methodology and artificial neural network analysis to optimize the MAA extraction process from P. yezoensis. Based on single-factor experimental investigations, response surface and artificial neural network models were designed and constructed using the MAA extraction rate as an indicator, followed by a comparative analysis and validation of the two modeling approaches. These findings suggest that both models consistently demonstrate that the solid-liquid ratio and extraction time exert the most considerable influence on the extraction rate. The response surface model exhibited outstanding predictive capability and reliability, with the experimental extraction rate achieved under optimized conditions（0.468%） closely matching the predicted value（0.474%）. Although the artificial neural network model demonstrated high goodness-of-fit（R²>0.93） on the training and validation sets, it exhibited substantial prediction errors on the independent test set. This finding suggests that the model exhibits inferior generalization capability in comparison with the response surface model. The optimal extraction conditions were determined to be as follows: temperature: 42 ℃, time duration: 1.51 h, solid-to-liquid ratio: 1∶25（g/mL）, extraction cycles: 2 cycles, and methanol concentration: 20%. Under these conditions, the extraction yield was found to be 0.468%. This study provides a valuable reference point for the large-scale extraction and utilization of MAAs from P. yezoensis.

Key words: Mycosporine-like amino acids Pyropia yezoensis extraction process response surface artificial neural network