吲唑类PI3Kδ选择性抑制剂活性理论模型的构建

Construction of the Theoretical Model for Predicting the Activity of Indazole Selective Inhibitors of PI3Kδ

吲唑类PI3Kδ选择性抑制剂是一类新的靶向抗癌药物,有的已开始应用于临床.为了研究PI3Kδ抑制剂的活性pIC50与结构之间的定量结构-活性关系,在分子拓扑理论的基础上,计算了47个PI3Kδ抑制剂分子的分子形状指数和电性拓扑状态指数,筛选了分子形状指数的K1,K3,K4和电性拓扑状态指数的E7,E16,E19共6种结构参数,建立了这些PI3Kδ抑制剂分子的活性与K1,K3,K4,E7,E16,E19这6种结构参数的定量结构-活性相关性多元回归模型,并将这6种参数作为神经网络输入层变量,47个PI3Kδ抑制剂分子的活性为输出层变量,采用6∶4∶1的网络结构方式,建立了预测能力强的神经网络模型,其总相关系数r为0.9794,利用该神经网络模型计算得到的预测活性值与实验值的平均相对误差为1.92%.结果表明,PI3Kδ选择性抑制剂的活性与6种分子结构参数之间有良好的非线性关系,基团的类型、基团的连接位置以及相互之间的作用均能影响PI3Kδ选择性抑制剂的活性.

Indazole selective inhibitors of PI3Kδ were new types of targeted anticancer drugs, and some of them have been applied in clinical practice. In order to study the quantitative structure-activity relationship (QSAR) between the activity pIC 50 and the molecular structure of PI3Kδ inhibitors, the molecular shape indices and electrotopological state indices of 47 PI3Kδ inhibitors were calculated based on the molecular topological theory. The molecular shape indices K 1, K 3, K 4 and electrotopological state indices E 7, E 16 , E 19 were screened. The QSAR multivariate regression model was developed based on the activity pIC 50 and six parameters ( K 1, K 3, K 4, E 7, E 16 , E 19 ) of 47 PI3Kδ inhibitors. Then, the six structural parameters were used as input layer variables, and the activity pIC 50 of 47 PI3Kδ inhibitors were used as output layer variable of the neural network. A neural network model with strong predictive ability was established and the neural network structure was 6∶4∶1. The total correlation coefficient r was 0.979 4. The relative average error between the experimental values and the predicted values of activity pIC 50 was 1.92%. The results show that there is good nonlinear relationship between the activity pIC 50 and the six molecular structure parameters. The type of group, the joining position of group and the interaction between them all can effect the activity of selective inhibitor of PI3Kδ.