FT-NIR光谱法快速预测烟草中的游离氨基酸

Rapid prediction of contents of free amino acids in tobacco by Fourier transform-near infrared spectroscopy

为研究应用傅里叶变换近红外光谱（FT-NIR）法快速测定烟草中游离氨基酸质量分数的可行性，采用偏最小二乘法（PLS）作为建模方法，选择4000～9000cm^-1谱段，采用一阶导数和Norris滤波法进行光谱预处理，建立了烟草中的组氨酸、天冬酰胺、谷氨酰胺、脯氨酸、苯丙氨酸和色氨酸以及游离氨基酸总量的近红外预测模型；对比了近红外光谱技术和常规标准检测方法，并验证了模型偏差的显著性。结果表明：①6种氨基酸以及游离氨基酸总量的最佳PLS预测模型的相关系数（r）均大于0．9，预测标准偏差（SEP）均小于校正标准偏差（SEC）的1．2倍，验证集样本的标准偏差（SD）与SEP的比值均大于2。②烟草中的组氨酸、天冬酰胺、谷氨酰胺、脯氨酸、苯丙氨酸和色氨酸以及游离氨基酸总量校正模型的交叉验证均方差（RMSECV）分别为0．0217、0．1860、0．2020、0．8180、0．0326、0．0478和1．5920。③在显著水平0．05时，两种方法不存在显著性差异。近红外光谱技术可以准确地定量分析烟草中的6种游离氨基酸以及游离氨基酸的总量。

In order to study the feasibility for a rapid method to calculate the contents of free amino acids in tobacco by Fourier transform near-infrared （FT-NIR） spectroscopy, the NIR spectroscopy models for calculating the contents of histidine （His）, asparagine （Ash）, glutamine （Gln）, proline （Pro）, phenylalanine （Phe）, tryptophan （Trp） and the total content of free amino acids in tobacco were developed. The approach used a partial least squares （PLS） regression method to model and the first derivative and Norris filter to pretreat obtained spectra in the range of 4 000-9 000 cm^-1. The results from the NIR spectroscopy method and conventional standard method were compared, and the significance of deviation of the developed models was verified. The results showed that： 1） For the optimal PLS prediction models of the said amino acids and the total content of the free amino acids, the correlation coefficients （r） were higher than 0.9, the ratios of standard error of prediction （SEP） to standard error of calibration （SEC） were less than 1.2, and the ratios of standard deviation （SD） of verification set samples to SEP were higher than 2. 2） The root mean square error of cross validation （RMSECV） of the seven calibration models for the contents of the selected amino acids and the total content of free amino acids were 0.021 7, 0.186 0, 0.202 0, 0.818 0, 0.032 6, 0.047 8 and 1.592 0, respectively. 3） There was no significant difference between the two methods at the significant level of 0.05. Therefore, the NIR spectroscopy technology can be used to accurately quantitatify the contents of these amino acids and the total content of free amino acids in tobacco.