摘要
荞麦籽粒中富含谷类作物比较缺乏的赖氨酸,使其不同于其他作物,具有较高经济价值。传统氨基酸测定费时且昂贵,为满足金苦荞育种工作的需要,选用近红外光谱技术结合人工神经网络的算法建立快速检测金苦荞叶片中氨基酸含量的近红外模型。使用氨基酸含量差异较大的样品255个,扫描光谱后测定其化学值。研究发现样品苏氨酸(Thr)含量范围是5.307~14.374 mg·g^(-1);缬氨酸(Val)含量范围是6.137~16.204 mg·g^(-1);甲硫氨酸(Met)含量范围是0.308~3.049 mg·g^(-1);异亮氨酸(Ile)含量范围是5.259~14.134 mg·g^(-1);亮氨酸(Leu)含量范围是9.730~26.061 mg·g^(-1);苯丙氨酸(Phe)含量范围是5.936~17.223 mg·g^(-1);赖氨酸(Lys)含量范围是6.640~17.280 mg·g^(-1);谷氨酸(Glu)含量范围是10.984~27.740 mg·g^(-1);天冬氨酸(Asp)含量范围是6.437~17.280 mg·g^(-1);丝氨酸(Ser)含量范围是3.467~8.312 mg·g^(-1);精氨酸(Arg)含量范围是4.937~14.772 mg·g^(-1);丙氨酸(Ala)含量范围是3.329~6.885 mg·g^(-1);组氨酸(His)含量范围是1.946~4.798 mg·g^(-1);甘氨酸(Gly)含量范围是4.196~9.264 mg·g^(-1);脯氨酸(Pro)含量范围是1.024~5.672 mg·g^(-1);酪氨酸(Tyr)含量范围是0.176~1.173 mg·g^(-1);半胱氨酸(Cys)含量范围是0.422~1.926 mg·g^(-1)。每次随机选取50个样品建设模型,以4∶1的比例随机分为训练集和测试集。数据进行归一化处理后,使用神经网络结构1102-9-1进行模型建设。利用多次学习的方式建立了较优模型,其中Arg和Asp近红外模型的仿真测试结果最好,预测值与真实值的相关系数(R^(2))均大于0.97,平均相对误差(RSD)也小于10%;另外Leu,Val,Tyr,Ile,Ser,Ala,Thr,His,Phe,Gly和Lys模型的R^(2)均大于0.90,模型仿真测试数据的RSD小于10%,模型均可用;Met与Cys的模型进行仿真测试时,其预测值与真实值的R^(2)均大于0.78,但RSD大于10%,模型不可用。结果表明,金苦荞叶片的氨基酸含量高,有极高应用价值,近红外光谱技术结合人工神经网络的分析方法可应用于金苦荞氨基酸含量的预测,为高品质荞麦育种工作提供了便利。
Buckwheat is rich in lysine,which is a lack in cereal crops,making it different from other cereal crops and has high economic value.Traditional determination of amino acids was time-consuming and expensive.In order to meet the needs of breeding of golden Tartary buckwheat,this study selected near-infrared spectroscopy combined with an artificial neural network algorithm to establish a near-infrared model for rapid determination of amino acid content in buckwheat leaves.A total of 255 samples with different amino acid contents were studied,and their chemical values were determined after scanning spectra.It was found that the content of threonine(Thr)in the samples ranged from 5.307 to 14.374 mg·g^(-1).Valine(Val)content ranged from 6.137 to 16.204 mg·g^(-1).The content of methionine(Met)ranged from 0.308 to 3.049 mg·g^(-1).The content of isoleucine(Ile)ranged from 5.259 to 14.134 mg·g^(-1).Leucine(Leu)content ranged from 9.730 to 26.061 mg·g^(-1).The content of phenylalanine(Phe)ranged from 5.936 to 17.223 mg·g^(-1).Lysine(Lys)content ranged from 6.640 to 17.280 mg·g^(-1).The content of glutamic(Glu)ranged from 10.984 to 27.740 mg·g^(-1).Aspartic(Asp)content ranged from 6.437 to 17.280 mg·g^(-1).Serine(Ser)content ranged from 3.467 to 8.312 mg·g^(-1).Arginine(Arg)content ranged from 4.937 to 14.772 mg·g^(-1).The content of Alanine(Ala)ranged from 3.329 to 6.885 mg·g^(-1).Histidine(His)content ranged from 1.946 to 4.798 mg·g^(-1).The content of glycine(Gly)ranged from 4.196 to 9.264 mg·g^(-1).Proline(Pro)content ranges from 1.024 to 5.672 mg·g^(-1).The content of tyrosine(Tyr)ranged from 0.176 to 1.173 mg·g^(-1).The content of cysteine(Cys)ranged from 0.422 to 1.926 mg·g^(-1).During each modeling,50 samples were randomly selected and randomly divided into the training set and test set at a ratio of 4∶1.After data normalization,the neural network structure 1102-9-1 was used for model construction.The simulation results of Arg and Asp near-infrared models were the best,the correlation coefficient(R^(2))between the predicted value and the real value was greater than 0.97,and the average relative error(RSD)was less than 10%.Simulation test process found,Val,Tyr,Ile,Ser,Ala,Thr,His,Phe,Gly and Lys of model prediction and the real value of R^(2) are greater than 0.90,the RSD is less than 10%,models are available;When the models of Met and Cys were tested in simulation,the R^(2) between the predicted value and the true value were both greater than 0.78,but the RSD was greater than 10%,so the model was not available.The results showed that golden Tartary buckwheat leaves had a high content of essential amino acids and had high application value.The analysis method of near infrared spectroscopy combined with an artificial neural network could be used to predict the amino acid content of buckwheat,which provided convenience for the breeding of high-quality buckwheat.
作者
朱丽伟
严金欣
黄娟
石桃雄
蔡芳
李洪有
陈庆富
陈其皎
ZHU Li-wei;YAN Jin-xin;HUANG Juan;SHI Tao-xiong;CAI Fang;LI Hong-you;CHEN Qing-fu;CHEN Qi-jiao(Research Center of Buckwheat Industry Technology,Guizhou Normal University,Guiyang 550001,China)
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2022年第1期49-55,共7页
Spectroscopy and Spectral Analysis
基金
国家自然科学基金项目(31760430)
国家重点研发计划项目(2019YFD1001300,2019YFD1001304)
贵州省科技计划项目(黔科合平台人才[2017]5726-20,黔科合基础[2016]1106号)
贵州师范大学博士启动基金项目(11904/0516027)
贵州省荞麦种质资源保育及创新重点实验室建设专项基金项目(黔教合KY字[2017]002)资助。
关键词
荞麦
近红外
人工神经网络
氨基酸
模型
Buckwheat
Near infrared spectroscopy
Artificial neural network
Amino acid
Model