大豆分离蛋白酶解液抗氧化性的近红外光谱定量测定

利用近红外光谱技术对四种蛋白酶(菠萝、碱性、木瓜和中性蛋白酶)的大豆分离蛋白水解样品进行测定,探索同一模型应用于不同酶的水解液抗氧化性测定的可行性。基于留一交叉验证方法,分析了采样密度和酶物质差异对全波长模型精度的影响,统计检验表明,采样密度与模型精度之间正相关,酶物质差异对单酶样本模型精度无显著影响,但对综合酶样本模型性能有显著影响。为了建立能够同时准确测定不同酶水解样品抗氧化性能力的综合酶样本预测模型,竞争性自适应重加权抽样(CARS)法用来对模型进行了优化,建模后,校正集R_(cv)和RMSECV分别为0.960 1和0.002 8,验证集的R_p和RMSEP为0.923 7和0.005 3,相对分析误差(RPD)为2.45,预测精度较好,说明建立针对不同酶的水解样品的同一模型是可行的。

Quantitative Determination of Antioxidant Activity of Hydrolysates from Soy Protein Isolate by Near Infrared Spectroscopy Combined with CARS

The hydrolyzed samples of soy protein isolated from four kinds of protease (pineapple, alkaline, papaya and neutral protease) were determined by near infrared spectroscopy and the feasibility of the same model applied to the determination of antioxidant activity of different enzymes was explored. Based on the method of left one cross validation, the influence of sampling density and enzyme substance on the accuracy of full wavelength model was analyzed. The statistical results showed that there was a positive correlation between sampling density and model accuracy. The difference of enzyme material had no significant effect on the accuracy of single enzyme sample model, but it had a significant influence on the performance of the integrated enzyme sample model. In order to establish a comprehensive model of enzyme prediction for simultaneous determination of the antioxidant capacity of different enzymes, the competitive adaptive weighting sampling (CARS) method was used to optimize the model. After modeling, Rcv and RMSECV of the calibration set were 0.960 1 and 0.002 8. Rp and RMSEP of the validation set were 0.923 7 and 0.005 3, and the relative analysis error (RPD) was 2.45. The result showed that the prediction accuracy was good, indicating that it is feasible to establish the same model for hydrolyzing the samples of different enzymes.