基于太赫兹光谱的转基因大豆及豆油检测机理研究

It's well known that both genetically modified Roundup Ready soybeans and its products contain insect resistant gene. Currently, there is a huge controversy in the domestic and foreign academic circles about whether it is harmful or not to human health, and its biological safety remains inconclusive. Therefore, the establishment of accurate and effective detection methods becomes the current most urgent task. In this project, for the first time, the spectroscopy technology is introduced to the qualitative identification and quantitative detection of transgenic soybean and its products, soybean oil and soybean meal..The main research contents of this project are summarized as follows. (1) According to the characteristics of terahertz wave, a new signal analysis algorithm, Clifford algebra, is proposed. This proposed algorithm may reveal the corresponding relationship between the optical parameters of the measured object and its terahertz spectrum. That research will provide some new ideas on the substance identification based on the terahertz spectrum. (2) Based on the Clifford algebra analysis method, the intrinsic absorption characteristics of CP4-EPSPS protein are extracted from soybean absorption spectrum, and the qualitative identification method of transgenic RR soybean and its parents based on CP4-EPSPS protein characteristic spectrum is explored. (3) Utilize many fields knowledge of spectroscopy and chemo-metrics to study on the qualitative identification mechanism of transgenic soybean oil. (4) Based on their spectral characteristics of, some research about how to quantitative detect the proportion of transgenic soybean in soybean products will be achieved. (5) Study the quantitative detects mechanism of the transgenic soybean oil proportion in the transgenic and non-transgenic mixing soybean oil. (6)At last but not the least, the qualitative and quantitative model for the THz spectrum is optimized to solve the problem of small-sample prediction and improve the generalization performance of the model while ensuring the steady-state accuracy.

转基因RR大豆及其制品中含抗虫基因，对人体是否有害即生物安全性仍无科学定论，故建立有效的检测方法非常迫切。本课题首次将太赫兹光谱技术引入到转基因大豆、豆油的定性鉴别、定量检测。主要研究内容包括：(1)引入Clifford代数分析方法，揭示被测对象光学参数与太赫兹信号之间存在的对应关系，探索基于太赫兹光谱的物质识别机理；(2)结合Clifford代数分析方法，从大豆吸收谱中提取CP4-EPSPS蛋白本征吸收特性，探索基于CP4-EPSPS蛋白特征光谱的转基因RR大豆与其亲本的定性鉴别方法；(3)基于太赫兹光谱的转基因与非转基因豆油化学计量学定性鉴别机理研究；(4)基于太赫兹光谱的豆油光谱与大豆混合比例的定量检测机理研究；(5)基于太赫兹光谱的转基因RR豆油与非转基因豆油混合比例定量检测机理研究；(6)对THz定性、定量模型算法优化，解决小样本预测的问题，在确保稳态精度的同时提高模型泛化性能。

本项目完成了预定研究内容，基本实现预期研究目标。在该项目资助下发表论文18篇，其中SCI、EI检索论文9篇，SCI检索5篇，EI检索1篇。本项目主要研究工作及其结论如下：.1.系统分析太赫兹（THz）信号产生、传输、探测和光学常数提取过程中引入的系统误差和随机误差，利用密度泛函理论(DFT)，基于6-31G*(d，p)基组对酒石酸THz吸收谱进行理论模拟计算与实验测量谱对照分析，探索采用自举软缩减法、竞争性自适应加权采样、蒙特卡洛无信息变量消除法及间隔区间偏最小二乘法对THz频域信号进行分离，提取目标成分光谱本征吸收特性。.2.收集不同年份，国、内外转基因、非转基因大豆样本共计252份，测定水分(%)、酸价(mg/g)、过氧化值(mmol/kg)、蛋白含量(%)、维生素E(mg/100g)等指标，为后续光谱信号的对照分析提供了必要的支持。将太赫兹光谱与化学计量学算法相结合，建立的转基因和非转基因大豆鉴别模型，鉴别准确率达96.49%，经过优化选取谱区范围和均值中心化预处理后，进一步提高转基因鉴别模型的准确率至98.25%。在此基础上进一步实现了基于THz吸收光谱的大豆中蛋白质、酸价、过氧化值维生素E定量检测，及样品来源产地鉴别。.3.收集、加工转基因、非转基因豆油87份，按大豆原料比重、豆油体积比配制不同比例豆油样本200份，进行THz透射光谱定量分析，样本太赫兹吸收系数随着转基因大豆油比例增加而呈现升高的趋势，体现出样本中转基因豆油比例与样本吸收系数幅值正相关。.4.对THz模型维护、传递及更新技术进行研究，采用光谱空间转换技术，实现了THz光谱定量模型在T-ray 5000与T-Spec光谱系统间的传递，基于该方法将主、从设备间模型预测均方根误差分别从1.5402、1.4884降至1.1516、0.9106，显著提升检测模型在不同检测设备、检测环境下适用性、泛化性。