桉树育种和遗传分析是开展桉树世代改良及其目标性状改良等研究的前提。 而常用的遗传基础研究方法专业性要求高, 且费时费力。 该研究旨在利用近红外光谱（NIRs）分析NIRs信息与桉树遗传信息间的关系, 并探索NIRs信息用于桉树杂交种判别分析的可行性和准确性。 以现有的桉树杂交种测试试验及其亲本材料为对象, 用手持式近红外仪Phazir Rx（1624）采集了7个桉树杂交种及其4个亲本树种叶片的NIRs信息。 每个树种选择10个单株, 每个单株选10片当年生健康叶片, 扫描其正面叶脉中部两侧光谱各5次, 以均值代表单个叶片的NIRs信息。 每种基因型总共各获得100条NIRs信息, 其中70条构成训练集样本, 30条构成验证集样本。 原始NIRs信息采用S.G二阶导数转换预处理, 以消除基线及其他因素对光谱信息的影响, 增强特征峰信息。 经预处理后的NIRs信息用于后续分析, 首先通过主成分分析（PCA）的因子得分对树种的分类判断NIRs信息与测试树种遗传信息间的关系。 在此基础上, 分别用簇类独立软模式（SIMCA）和偏最小二乘判别分析（PLS-DA）两种判别模式建立桉树杂交种的NIRs判别模型。 经预处理后的NIRs信息的变异系数曲线显示, 在波长2 000 nm后, 各树种的NIRs信息存在丰富的特征峰, 且特征峰的分布范围存在较大的差异。 PCA结果显示, 不同的亲本间、 杂交种间及杂交种与亲本间样本的PC1和PC2得分可以清晰地将各树种进行分类, 这在很大程度上表明NIRs信息可以正确反映桉树不同基因型的遗传信息。 NIRs模型的判别效果显示, 少数遗传关系比较接近的杂交组合的SIMCA模式相互判别准确率较低, 而多数杂交组合间的SIMCA判别准确率则在73%～100%之间; 桉树各杂交组合间的单独和综合模型的PLS-DA判别准确率均为100%, 且基于PLS-DA判别的综合模型能将7个杂交组合一一与其他组合正确区分开, 判别效果明显优于SIMCA模式。 结果表明: NIRs信息可以正确反映桉树不同基因型的遗传信息, NIRs判别模型可以比较准确地将各树种进行区分, 因此, NIRs信息可用于桉树杂交种和纯种的田间定性判别, 从而辅助桉树育种材料遗传基础的研究。

The analysis of genetic basis of breeding materials is the precondition for the improvement programs on populations and interesting traits in eucalypt. However, the traditional ways for that have high professional requirements and are time-consuming and laborsome. The aim of present study was to study the relationship between NIRs and genetic information of eucalypt, and discuss the practicability and the accuracy of the discriminant model for the classification of eucalypt hybrids by NIRs data. The NIRs of seven eucalypt hybrids and four parental pure species were scanned with healthy leaves using handheld portable near infrared spectrometer Phazir Rx (1624). 10 individuals were selected for a genotypic species, and 10 healthy current-year leaves were chosen per individual tree. Specially five scans for NIRs from each side of the middle part of the frontal vein of the leaves were taken, and estimated the average of that as the NIRs information of a leaf. In total, 100 NIRs were gained per genotypic species, 70 of which constitute the calibration set, and the validation set consists of the rest 30 NIRs. The transformation of S.G 2nd derivative were performed for the raw NIRs data in present study so as to eliminate the effects of baseline and other factors on the NIRs information, and to strengthen the characteristic peaks of NIRs. The later analysis were conducted after the pretreatment. Firstly, the relationship between NIRs and genetic information of eucalypt hybrids was studied by the scores plot of principal components (PCs) in principal component analysis (PCA), and on this basis, the NIRs discriminant model was developed. The soft independent modeling of class analogy (SIMCA) and partial least squares-discriminant analysis (PLS-DA) pattern recognition were used to classify eucalypt hybrids with the NIRs model calibrated. The coefficient variation curves of NIRs transformation showed that all phenotypic species studied had rich characteristic peaks, and big differences among them after the wavelength of 2000 nm. The scores plot of PC1 and PC2 in PCA demonstrated clear groups among parental species, hybrids, as well as between hybrids and their parents, suggesting NIRs was a direct response to the genetic information of different genotypes. The discriminant accuracy of SIMCA pattern recognition between some cross combinations, which shared close genetic relation of cross parents, were relatively low using NIRs model. In contrast, the discriminant accuracy of SIMCA pattern recognition among most of eucalypt combinations changed between 73% and 100%. The discriminant accuracy of PLS-DA pattern recognition using single and combined NIRs model of hybrids all were 100%, and the combined model of hybrids based on PLS-DA pattern can discriminate seven hybrids clearly. Studies showed that, the discriminant accuracy of PLS-DA pattern was much higher than that of SIMCA pattern recognition. The current study indicated that NIRs information is the correct response of different genotypic eucalypt species, and the NIRs calibrated model can classify different species of eucalypt accurately, so the NIRs would be used in the qualitative discrimination analysis of eucalypt hybrids and pure species in field, providing an alternative way for the analysis of genetic basis of breeding materials in eucalypt.