神经网络方法在血管紧张素转换酶抑制剂定量构效关系建模中的应用

Quantitative Structure-activity Relationship Modeling of Angiotensin Converting Enzyme Inhibitors by Back Propagation Artificial Neural Network

对20个ACEI化合物用量子化学方法进行结构优化并计算出10个参数,用9种不同隐含层节点数的BP神经网络研究了ACEI的定量构效关系,建立了节点为10/6/1的三层BP神经网络模型。结果表明:以量化理论计算所得参数可以构建合理的ACEI定量构效关系模型,神经网络模型M6的r2=0.995,S=0.050,6个验证集化合物的残差平方和为0.002,预测能力明显强于多元线形回归模型,亦优于同类文献报道,可作为ACEI研发领域中预测先导化合物活性的理论工具。

The geometries and electronic structures of 20 Angiotensin Converting enzyme inhibitors （ACEI） had been optimized by using quantum chemical methods and 10 quantum-chemical parameters such as energies, atom-charges etc. , were calculated. Nine back propagation （BP） artificial neural networks （ANN） with different nodes were trained to research quantitative structure-activity relationship （QSAR） of ACEI. A 3-layers BP artificial neural network model with 6 nodes in hidden layer was developed. The results of statistic analysis are ： r^2 = 0. 995, S = 0. 050, which shows that artificial neural network model M6 is more accurate and precise than multiple linear regression models. The established ANN model can be applied to predict the activity of ACEI trustfully.