虉草粗蛋白近红外定量分析模型的建立
|
摘要
虉草(Phalaris arundinacea L.)为多年生冷季型禾本科高产牧草,粗蛋白(CP)是评价饲草品质的关键指标,但目前的化学分析方法存在诸多缺点,寻求高效、快速、准确、安全的虉草CP测定方法是现代草地畜牧业发展和草原生态恢复急需解决的实际问题。本研究旨在利用近红外光谱(NIRS)建立虉草CP的定量分析模型,为快速测定虉草CP提供有效方法。试验采集不同品种(品系)、生育期、栽培条件、干燥方式、生长年限、部位以及刈割次数的虉草样品454份,采用瑞士Buchi公司的傅里叶近红外光谱仪和Operator软件采集原始光谱,应用K-S算法剔除具有相似光谱的样品,筛选出210份用于建模和模型评价。通过凯氏定氮法测定210份样品的粗蛋白含量并在Management console软件中对光谱进行赋值,再采用软件NIRcal 5.4按照6∶3的比例将样品随机分为校正集和验证集,并剔除异常样品,运用不同的光谱预处理、回归算法、建模波段和主成分数建立8个虉草CP含量的近红外定量分析模型,通过外部验证表明8个模型均可以进行实际测定。最后比较不同的统计学参数获取最佳模型。结果表明,采用4 000～10 000 cm~(-1)的建模光谱波段、 sa3+ncl+db1(3点平滑+趋近归一化+一阶导数处理)的预处理方法、 8/1-4的初/次级主成分数和偏最小二乘法(PLS)所建的模型为最佳模型,其校正决定系数(R_(cal)~2)为0.982 1,验证决定系数(R_(val)~2)为0.980 2,均大于0.98,表明预测性能优秀;校正标准差(SEC)和验证标准差(SEP)分别为0.780 2和0.783 2,均较小且非常接近,表明模型的分析精度很高并具有很好的适应性;残差(BIAS)为-0.000 5,接近于0,说明模型的稳定性很高,对外界因素不敏感;预测相关系数(r)为0.99,可见样品化学值与定标模型预测值的相关度极高;相对分析误差(RPD)为7.37, RPD>4.0表明模型能够很好地进行定量分析。综上,该试验在国内首次建立了虉草CP近红外定量分析模型,该模型样品来源多、数量大、分布范围广,预测精度和准确度高,适用范围大,为快速测定虉草粗蛋白提供了有效方法,在虉草品质分析、育种、家畜日粮配置以及草产品评价流通等方面具有应用前景。
Reed canary grass(Phalaris arundinacea L.) is a perennial cool-season gramineae grass with a high yield. Crude protein(CP) is a key indicator in the evaluation of forage quality, but the use of chemical analytical methods to determine the CP content is disadvantageous. Therefore, a fast, efficient, accurate, and safe determination method is required in the development of modern grassland agriculture, animal husbandry and grassland ecological restoration. The purpose of this study is to use near-infrared spectroscopy(NIRS) techniques to develop a quantitative model for the analysis of CP in reed canary grass and provide an effective method for a rapid determination. We collected 454 samples of reed canary grass from various resources, including different cultivars(or strains), different growth stages, different cultivation conditions, different drying methods, different growth years, different parts and different harvest times. The original spectra of all of the samples were obtained using a near-infrared spectrometer(NIRFlex N-500) and Operator software of the Swiss Buchi company. A total of 210 samples were selected for the development and evaluation of models after deleting samples with similar spectra by a K-S algorithm, and were assayed using the Kjeldahl nitrogen method to obtain the chemical values of CP; we then assigned them to spectra in a Management console software. The samples were randomly divided into a calibration set and a validation set according to the proportion of 6∶3, using the NIRcal 5.4 software; the outliers were then eliminated. We established 8 quantitative analysis models for the CP content of reed canary grass by applying different spectral data pretreatments, primary/secondary principal components, spectral regions, and regression algorithms. We revealed that all of the 8 models can be used in the determination of CP by performing an external validation. The best model was chosen by comparing statistic parameters. The results showed that the best calibration model was developed by the spectral data pretreatment of sa3+ncl+db1(smoothing average 3 points+ normalization by closure+first derivative BCAP), choosing the primary/secondary principal component of 8/1-4 and spectral region of 4 000～10 000 cm~(-1) in combination with the partial least square(PLS) regression algorithm. Its calibration coefficient of determination(R_(cal)~2) and external validation coefficient of determination(R_(val)~2) were 0.982 1 and 0.980 2, respectively; both were larger than 0.98, suggesting an excellent predictive ability. The standard errors of calibration(SEC) and prediction(SEP) were 0.780 2 and 0.783 2, respectively; both were very small and similar, which demonstrated the high analytical accuracy and robust fitting. The bias value of-0.000 5, close to 0, demonstrated the model's stability and robustness, i. e., its insensitivity to the external factors. The correlation coefficient of validation(r) of 0.99 indicated a very high correlation between the predicted and chemical values. The residual predictive deviation(RPD) was 7.37(above 4.0), further confirming that the CP model can be used for a high-quality quantitative analysis. Therefore, in this study, a quantitative model for a CP analysis of Phalaris arundinacea L. was developed using NIRS for the first time in China with a large data collection from different sources and high accuracy, which guaranteed the reliability and practicability. The model provides an effective method to quantify CP of reed canary grass for a rapid screening of germplasm in breeding programs, optimization of the allocation of livestock diets, and classification of forage products in the supply chain.
Reed canary grass(Phalaris arundinacea L.) is a perennial cool-season gramineae grass with a high yield. Crude protein(CP) is a key indicator in the evaluation of forage quality, but the use of chemical analytical methods to determine the CP content is disadvantageous. Therefore, a fast, efficient, accurate, and safe determination method is required in the development of modern grassland agriculture, animal husbandry and grassland ecological restoration. The purpose of this study is to use near-infrared spectroscopy(NIRS) techniques to develop a quantitative model for the analysis of CP in reed canary grass and provide an effective method for a rapid determination. We collected 454 samples of reed canary grass from various resources, including different cultivars(or strains), different growth stages, different cultivation conditions, different drying methods, different growth years, different parts and different harvest times. The original spectra of all of the samples were obtained using a near-infrared spectrometer(NIRFlex N-500) and Operator software of the Swiss Buchi company. A total of 210 samples were selected for the development and evaluation of models after deleting samples with similar spectra by a K-S algorithm, and were assayed using the Kjeldahl nitrogen method to obtain the chemical values of CP; we then assigned them to spectra in a Management console software. The samples were randomly divided into a calibration set and a validation set according to the proportion of 6∶3, using the NIRcal 5.4 software; the outliers were then eliminated. We established 8 quantitative analysis models for the CP content of reed canary grass by applying different spectral data pretreatments, primary/secondary principal components, spectral regions, and regression algorithms. We revealed that all of the 8 models can be used in the determination of CP by performing an external validation. The best model was chosen by comparing statistic parameters. The results showed that the best calibration model was developed by the spectral data pretreatment of sa3+ncl+db1(smoothing average 3 points+ normalization by closure+first derivative BCAP), choosing the primary/secondary principal component of 8/1-4 and spectral region of 4 000～10 000 cm~(-1) in combination with the partial least square(PLS) regression algorithm. Its calibration coefficient of determination(R_(cal)~2) and external validation coefficient of determination(R_(val)~2) were 0.982 1 and 0.980 2, respectively; both were larger than 0.98, suggesting an excellent predictive ability. The standard errors of calibration(SEC) and prediction(SEP) were 0.780 2 and 0.783 2, respectively; both were very small and similar, which demonstrated the high analytical accuracy and robust fitting. The bias value of-0.000 5, close to 0, demonstrated the model's stability and robustness, i. e., its insensitivity to the external factors. The correlation coefficient of validation(r) of 0.99 indicated a very high correlation between the predicted and chemical values. The residual predictive deviation(RPD) was 7.37(above 4.0), further confirming that the CP model can be used for a high-quality quantitative analysis. Therefore, in this study, a quantitative model for a CP analysis of Phalaris arundinacea L. was developed using NIRS for the first time in China with a large data collection from different sources and high accuracy, which guaranteed the reliability and practicability. The model provides an effective method to quantify CP of reed canary grass for a rapid screening of germplasm in breeding programs, optimization of the allocation of livestock diets, and classification of forage products in the supply chain.
引文
[1] Zhang C,Zhang D W,Sun Y N,et al.Journal of Plant Physiology,2017,215:11.
[2] ZHANG Yong-liang,LUO Xiu-mei(张永亮,骆秀梅).Acta Agrestia Sinica(草地学报),2008,16(6):659.
[3] Ye D,Sun L,Zou B,et al.Spectrochimica Acta Part A:Molecular and Biomolecular Spectroscopy,2018,189:463.
[4] López A,Arazuri S,Jarén C,et al.Procedia Technology,2013,8(1):488.
[5] Ercioglu E,Velioglu H M,Boyaci I H.Talanta,2018,178:716.
[6] YAN Xu,BAI Shi-qie,YAN Jia-jun,et al(严旭,白史且,鄢家俊,等).Spectroscopy and Spectral Analysis(光谱学与光谱分析),2015,35(8):2103.
[7] Magwaza L S,Naidoo S I M,Laurie S M,et al.LWT-Food Science and Technology,2016,72:63.
[8] CHEN Guang-ji,SONG Shan-dan,GUO Chun-hua,et al(陈光吉,宋善丹,郭春华,等).Acta Prataculturae Sinica(草业学报),2015,24(9):63.
[9] JIA You-yu,LUO Xiu-mei,ZHANG Yong-liang(贾有余,骆秀梅,张永亮).Pratacultural Science(草业科学),2016,33(11):2312.
[10] LUO Xiu-mei,ZHANG Yong-liang(骆秀梅,张永亮).Journal of Inner Mongolia University for the Nationalities·Natural Sciences(内蒙古民族大学学报·自然科学版),2017,32(5):418.
[11] Fagan C C,Everard Colm D,McDonnell K.Bioresource Technology,2011,102(8):5200.
[12] LIU Zhe,WANG Yu-qin,XUE Shu-yuan,et al(刘哲,王玉琴,薛树媛,等).Acta Agrestia Sinica (草地学报),2018,26(1):249.
[13] Gislum R,Mick lander E,Nielsen J P.Field Crops Research,2004,88:269.
[14] WANG Jia-hua,TANG Zhi-hui,HAN Dong-hai(王加华,汤智辉,韩东海).Journal of Food Safety and Quality(食品安全质量检测学报),2014,5(3):742.
[2] ZHANG Yong-liang,LUO Xiu-mei(张永亮,骆秀梅).Acta Agrestia Sinica(草地学报),2008,16(6):659.
[3] Ye D,Sun L,Zou B,et al.Spectrochimica Acta Part A:Molecular and Biomolecular Spectroscopy,2018,189:463.
[4] López A,Arazuri S,Jarén C,et al.Procedia Technology,2013,8(1):488.
[5] Ercioglu E,Velioglu H M,Boyaci I H.Talanta,2018,178:716.
[6] YAN Xu,BAI Shi-qie,YAN Jia-jun,et al(严旭,白史且,鄢家俊,等).Spectroscopy and Spectral Analysis(光谱学与光谱分析),2015,35(8):2103.
[7] Magwaza L S,Naidoo S I M,Laurie S M,et al.LWT-Food Science and Technology,2016,72:63.
[8] CHEN Guang-ji,SONG Shan-dan,GUO Chun-hua,et al(陈光吉,宋善丹,郭春华,等).Acta Prataculturae Sinica(草业学报),2015,24(9):63.
[9] JIA You-yu,LUO Xiu-mei,ZHANG Yong-liang(贾有余,骆秀梅,张永亮).Pratacultural Science(草业科学),2016,33(11):2312.
[10] LUO Xiu-mei,ZHANG Yong-liang(骆秀梅,张永亮).Journal of Inner Mongolia University for the Nationalities·Natural Sciences(内蒙古民族大学学报·自然科学版),2017,32(5):418.
[11] Fagan C C,Everard Colm D,McDonnell K.Bioresource Technology,2011,102(8):5200.
[12] LIU Zhe,WANG Yu-qin,XUE Shu-yuan,et al(刘哲,王玉琴,薛树媛,等).Acta Agrestia Sinica (草地学报),2018,26(1):249.
[13] Gislum R,Mick lander E,Nielsen J P.Field Crops Research,2004,88:269.
[14] WANG Jia-hua,TANG Zhi-hui,HAN Dong-hai(王加华,汤智辉,韩东海).Journal of Food Safety and Quality(食品安全质量检测学报),2014,5(3):742.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |