Centre International de Recherches Agronomiques pour le Développement, PO Box 946, Port-Vila, Vanuatu.
J Sci Food Agric. 2013 May;93(7):1788-97. doi: 10.1002/jsfa.6002. Epub 2012 Dec 18.
Thousands of yam (Dioscorea spp.) accessions are maintained in germplasm collections. The physico-chemical characteristics of the tubers are rarely characterised. Unless a simple, low cost, screening tool is available, it is difficult to evaluate the quality of varieties and breeding lines. We investigated the potential of near infrared reflectance spectroscopy (NIRS) as an alternative method for predicting the major constituents of the yam tuber.
Two hundred and sixty-five accessions, belonging to seven different Dioscorea spp., were analysed for starch, amylose, sugars, proteins, minerals and cellulose. The comparison of the NIR spectra and the chemical values allowed the establishment of equations of calibration for the prediction of starch, sugars and proteins (equivalent N). The r(2) pred values for starch, sugars and proteins (respectively 0.84, 0.86 and 0.88) are high enough to allow good estimates of their contents. Values for the ratio of performance to deviation (RPD) of 4.046 and 3.641 for the sugars and proteins models also allow good quantitative predictions to be made. Amylose, cellulose and minerals could not be predicted precisely. A second calibration conducted by adding the calibration and validation sets (260 accessions) revealed an improvement of the RPD values for starch, sugars and proteins, indicating that the models can be improved. Discriminant analysis conducted using 2151 wavelengths (in nanometres) as variables was applied to a set of 214 accessions of D. alata and the results were compared to the principal component analysis of chemical data. Accessions can be classified according to the amylaceous fraction of the chemotype.
NIRS could be used in yam breeding programmes to characterise rapidly and at low cost the numerous accessions and breeding lines.
成千上万的山药（薯蓣属）品种被保存在种质资源库中。薯蓣块茎的物理化学特性很少被描述。除非有一种简单、低成本的筛选工具，否则很难评估品种和品系的质量。我们研究了近红外反射光谱（NIRS）作为替代方法来预测山药块茎主要成分的潜力。
对 265 个属于七个不同薯蓣属的品种进行了淀粉、直链淀粉、糖、蛋白质、矿物质和纤维素分析。NIR 光谱和化学值的比较允许建立预测淀粉、糖和蛋白质（等效 N）的校准方程。淀粉、糖和蛋白质的 r(2)预估值（分别为 0.84、0.86 和 0.88）足够高，可以很好地估计它们的含量。糖和蛋白质模型的性能偏差比（RPD）值分别为 4.046 和 3.641，表明可以进行良好的定量预测。直链淀粉、纤维素和矿物质不能被精确预测。通过添加校准和验证集（260 个品种）进行的第二次校准显示，淀粉、糖和蛋白质模型的 RPD 值有所提高，表明模型可以得到改进。使用 2151 个波长（纳米）作为变量进行的判别分析应用于一组 214 个 D. alata 品种，结果与化学数据的主成分分析进行了比较。可以根据品种的淀粉部分对品种进行分类。
NIRS 可用于山药育种计划，以快速、低成本的方式对大量的品种和品系进行特征描述。
