Determination of caseinomacropeptide in uht milk using near infrared spectroscopy and pls regression

ABSTRACT This work´s objective was to develop an UHT milk caseinomacropeptide determination method trough NIR spectroscopy application. Eight UHT milk trademarks are used to produce a mathematical calibration model. The results of NIR analysis suggested that to produce a suitable calibration model, partial least-square regression (PLSR) must be used, with maximum distance in wavelenght space to select spectra, pre - treatment with 2nd derivative and standard normal variant (SNV). Also, suitable near-infrared regions more correlated with CPM aminoacids vibrational movements: 1100-1310; 1400-1430; 1490-1550; 1640-1680; 1780-1970; 2020-2100; and 2310-2350nm. Therefore, NIR spectroscopy can be an alternative to caseinomacropeptide determination of UHT milk, since there was a representative calibration set with a large enough and representative sample of entire population to be predicted in the future.

摘要 本研究旨在开发一种基于近红外光谱（NIR, Near-Infrared Spectroscopy）的超高温灭菌（UHT, Ultra High Temperature）乳中酪蛋白巨肽（caseinomacropeptide）的检测方法。本研究选用8个UHT乳品牌样品构建数学校正模型。近红外光谱分析结果表明，构建性能优良的校正模型需采用偏最小二乘回归（PLSR, Partial Least-Square Regression）方法，结合波长空间最大距离法筛选光谱，并采用二阶导数与标准正态变换（SNV, Standard Normal Variant）进行光谱预处理。此外，筛选出与酪蛋白巨肽氨基酸振动模式相关性较高的近红外光谱区间：1100~1310nm、1400~1430nm、1490~1550nm、1640~1680nm、1780~1970nm、2020~2100nm以及2310~2350nm。综上，近红外光谱可作为UHT乳中酪蛋白巨肽检测的替代方法，这是因为校正集具备足够的样本量且能充分代表未来待预测的总体样本特征。