【作者】 于燕波；

【导师】 康定明；

【作者基本信息】 中国农业大学 ， 作物遗传育种， 2014， 博士

【摘要】 转基因生物良种培育是生物技术在农业领域的重大突破与应用,对于保证我国粮食安全,积极应用高新技术促进农业生产发展是一项重要尝试。但是,由于转基因生物存在潜在的生物安全风险,需要加强和认真研究转基因生物的安全风险与管理措施,为此研发出快速方便和环境友好的转基因安全检测技术,对于推广应用和强化转基因生物的安全管理具有重要意义。水稻是我国的主要粮食作物,已有成功的转基因技术培育的水稻优良品种。本研究选用了来源于法国、美国、荷兰的TP309、中作321和日本晴三个品种六种转基因类型的47个转基因水稻材料,分别于2009年、2010年在中国农业大学上庄试验站种植,在水稻抽穗期检测叶片,并检测成熟的种子,获得了近红外漫反射光谱(叶片光谱530份,单粒种子样本光谱1500份,多种子样本光谱80份),对光谱数据进行了因子化法和定性偏最小二乘法的多品种、多转基因类型的转基因鉴定和基于(目标)主成分及Fisher准则的有监督模式的二重降维投影相似性分析。另外,论文还选用来源于不同棉花产区(包括陆地棉和海岛棉)棉花材料的棉籽145个,建立了棉籽含油量近红外定量测定模型,对棉籽含油量进行了相似性分析和定量分析相结合的高油棉籽筛选研究。主要研究结果如下：(1)采用DPLS方法建立了中作321、TP309和日本晴的水稻叶片、多种子样本、单粒种子样本近红外光谱品质预测模型,模型的平均识别正确率为86.51%、97.1%和91.55%,结果表明基于DPLS的近红外光谱分析技术可实现水稻品质的叶片鉴别、多种子样本鉴别和单粒种子鉴别。(2)采用DPLS方法建立了中作321、TP309和日本晴的水稻叶片和种子近红外光谱转基因识别模型。在单粒种子转基因识别中,中作321和日本晴转基因识别模型的识别正确率均达到100%,TP309的转基因识别模型识别正确率为93.4%,由此可见近红外光谱可实现水稻单粒种子样本的转基因识别。(3)对相同转基因事件的同种水稻单粒种子光谱进行相似性分析,研究发现转δ-OAT基因、P5CS基因的中作321与亲本对照差异明显,根据已有研究报道推测转δ-OAT基因、转P5CS基因的样本与亲本对照比较脯氨酸含量显著增加,部分农艺性状有变化。正是由于这种差异使得通过近红外技术识别转基因具有一定的可行性。对相同转基因不同插入位点的同种水稻种子进行转基因鉴别,发现插入不同位点TPS1基因的中作321基因表达性状差异不明显。(4)运用近红外漫反射技术研究快速预测水稻种子蛋白质和千粒重的方法,运用近红外漫反射技术研究快速、准确预测棉籽含油量的方法。所建棉籽含油量预测模型：分析谱区为8848.2cm-1～5442.4cm-1,光谱预处理方法为一阶导数+散射校正。模型采用内部交叉验证,建模集相关系数(r)为0.96,校正标准差(RMSECV)为1.13。对所建棉籽含油量预测模型进行外部验证,棉籽含油量预测模型检验集相关系数r高达0.98,预测相对误差小于5%。(5)运用近红外漫反射技术实现高油棉籽的无损筛选具有一定的可行性。通过PPF PCA方法建立了高油棉籽筛选的投影模型,实现有监督模式的二重降维,达到直观反映不同含油量的样品相似关系高低的目的,为高油棉籽的无损筛选提供了参考。更多还原

【Abstract】 GMO is a major breakthrough in the field of agriculture and Biotechnology. Positive application of high and new technology such as GMO seed cultivation to promote the development of agricultural production is an important attempt for our country to ensure food security. But because of GMO biological potential security risks, it is necessary to carefully study security risks and strengthen management. Therefore, this study developed rapid, convenient and environment friendly inspection technology for GMO. It is of great significance to the application and to strengthen safety management of GMO. Rice is the main food crops in China. breeding excellent rice variety by transgenic technology has been successful. Rice varieties selected in this study came from France, USA, and Holland. There were four transgenic types and47strains of TP309, Zhongzuo321and Nipponbare. They were planted in Shangzhuang experimental station of China Agricultural University in2009and2010. Detection of heading stage leaves and mature seeds by NIR, getting their near infrared diffuse reflectance spectra as the object to study (530leaves spectra,750single seed sample spectra and80multiple seed sample spectra). Using factorization method, DPLS and similarity analysis which were based on double principal component and supervised pattern Fisher criterion projection identificated varieties and transgenic. In addition,145cotton varieties from different cotton production areas (including upland cotton and Sea Island cotton), were chosed to establish the oil content N1R model. And by combinating determination and similarity analysis, the high oil cotton seed screening methods were developed.The major results were as follows:(1) Using DPLS built rice leaf, multiple seed sample and sigle seed sample NIR rice species prediction model for Zhongzuo321, TP309and Nipponbare.The correct rate of recognition model was86.51%,97.1%and91.55%. That means near infrared technology is able to be used identify the species of rice leaf, single seed and multiple seeds.(2) Using DPLS built rice leaf, mutiple seeds sample and sigle seed sample NIR transgenic identification model for Zhongzuo321, TP309and Nipponbare.The correct rate of sigle seed transgenic identification model for Zhongzuo321and Nipponbare were all100%, for TP309was93.4%. That means near infrared technology is able to be used identify the transgenic rice of single seed.(3) Using PPF projection studyed the same transgenic events of the same species. The study found the differences between Zhongzuo321(non transgenic) and P5CS genes, and the differences between Zhongzuo321(non transgenic) and δ-OAT genes. Some researches had reported that the samples with P5CS and δ-OAT gene were significantly high content of proline and were changed of some agronomic traits. It was because of this difference that made identifing transgenic by near infrared technology to be certain feasibility. The same transgenic insertion of TPS1gene in different sites couldn’t been identified by near infrared technique. That meant no significant differences in gene expression traits of TPS1gene in different sites.(4) Application of near infrared diffuse reflectance technique could quickly predict the protein content and thousand kernel weight of rice seed. The rapid and accurate models were established to predict cottonseed oil content by near infrared diffuse reflectance technique. Cottonseed oil content prediction model:The spectral region was8848.2cm-1～5442.4cm-1. The spectral pretreatment method was first-order derivative+MSC. The cross validation was used in building model. Its correlation coefficient (R) was0.96and its calibration standard deviation (RMSECV) was1.13. The coefficient of predictive cottonseed oil content and true oil content was up to0.98and the relative prediction error was less than5%.(5) It was of feasibility to screening high oil cottonseed nondestructively by near infrared diffuse reflectance technique. Projection model for screening high oil cottonseed was set up by PPF_PCA method, achieving double supervised pattern dimensionality reduction to reflect the similar of different oil content samples. It provided reference for high oil cottonseed screening.更多还原