基于神经网络的菠萝香气成分色谱保留值研究

为了研究菠萝香气成分色谱保留时间与其结构之间的定量构效关系,在分子拓扑理论基础上,计算了44种菠萝香气成分的分子价连接指数(mXtV)、分子形状指数(nK)和电拓扑状态指数(Ei)。优化筛选了分子价连接性指数2XVp和4XcV、分子形状指数1K和2K、电拓扑状态指数的E8和E13共6个参数,将其作为BP神经网络的输入层变量,香气成分的色谱保留时间作为输出层变量,采用6∶3∶1的网络结构,获得了令人满意的QSRR神经网络预测模型,模型总相关系数为0.995,计算得到的色谱保留时间的预测值与试验值吻合较好。结果表明,菠萝香气成分色谱保留时间与6种结构参数之间呈现良好的非线性关系,模型能较好地揭示香气成分色谱保留时间的递变规律。

Quantitative structure-retention relationship studies of aroma components from pineapple based on neural network

In order to study chromatographic retention time(RT)of aroma components from pineapple,molecular valence connectivity index(mXtV),molecular shape index(nK)and electrotopological state index(Ei)of 44 aroma components were calculated.2XpV and 4XcV of the molecular valence connectivity indices,and 1K and 2K of the molecular shape indices,E8 and E13 of the electrotopological state indices were optimized. The six parameters were used as input variables of neural network and the chromatographic retention time was used as output variable,and the 6∶3∶1 network structure was adopted and BP neural network method was used to establish a satisfying QSRR prediction model.The total correlation coefficient was 0.995.The predicted values by the model were in agreement those of the experiment values.A good nonlinear relationship between the chromatographic retention time and the six molecular structure parameters was found.The model could better elucidate the changing rule of chromatography retention time of the aroma components.