Unravelling effects of flavanols and their derivatives on acrylamide formation via support vector machine modelling
Unravelling effects of flavanols and their derivatives on acrylamide formation via support vector machine modelling
复制标题
通过支持向量机建模揭示黄烷醇及其衍生物对丙烯酰胺形成的影响
DOI：
10.1016/j.foodchem.2016.10.060
复制
发表时间：
2017-04-15
期刊：
影响因子：
8.8
通讯作者：
Zhang, Yu
中科院分区：
文献类型：
--
作者：
Huang, Mengmeng;Wang, Qiao;Zhang, Yu
本研究采用支持向量回归（SVR）方法预测黄烷醇及其衍生物对低水分条件下丙烯酰胺形成的影响。在马铃薯基等摩尔天冬酰胺-还原糖模型体系中，通过烤箱加热生成丙烯酰胺。黄酮类化合物在1 ~ 10000 μ mol/L处理范围内，对黄酮类化合物有正向和负向作用。黄烷醇及其衍生物（100 μ mol/L）对丙烯酰胺形成的抑制作用为59.9% ~ 78.2%，而在1万μ mol/L的对照中，其最大促进作用为2.15 ~ 2.84倍。抑制率与trolox当量抗氧化能力（Delta TEAC）变化的相关性（RTEAC-DPPH = 0.878, RTEAC-ABTS = 0.882, RTEAC-FRAP = 0.871）优于促进率（RTEAC-DPPH = 0.815, RTEAC-ABTS = 0.749, RTEAC-FRAP = 0.841）。以Delta TEAC为变量，优化后的SVR模型可以稳健性地作为预测效果的新工具（R: 0.783-0.880），其拟合性能略优于多元线性回归模型（R: 0.754-0.880）。(C) 2016 Elsevier Ltd.版权所有。
This study investigated the effect of flavanols and their derivatives on acrylamide formation under low-moisture conditions via prediction using the support vector regression (SVR) approach. Acrylamide was generated in a potato-based equimolar asparagine-reducing sugar model system through oven heating. Both positive and negative effects were observed when the flavonoid treatment ranged 1-10,000 mu mol/L. Flavanols and derivatives (100 mu mol/L) suppress the acrylamide formation within a range of 59.9-78.2%, while their maximal promotion effects ranged from 2.15-fold to 2.84-fold for the control at a concentration of 10,000 mu mol/L. The correlations between inhibition rates and changes in Trolox-equivalent antioxidant capacity (Delta TEAC) (RTEAC-DPPH = 0.878, RTEAC-ABTS = 0.882, RTEAC-FRAP = 0.871) were better than promotion rates (RTEAC-DPPH = 0.815, RTEAC-ABTS = 0.749, RTEAC-FRAP = 0.841). Using Delta TEAC as variables, an optimized SVR model could robustly serve as a new predictive tool for estimating the effect (R: 0.783-0.880), the fitting performance of which was slightly better than that of multiple linear regression model (R: 0.754-0.880). (C) 2016 Elsevier Ltd. All rights reserved.