羧酸类ALR2抑制剂生物活性的神经网络模拟研究  

A Neural Network Simulation Study on Biological Activity of Carboxylic Acid Targeting ALR2 Inhibitors

作　　者：陈艳[1] 李靖[1] 冯惠[1] CHEN Yan;LI Jing;FENG Hui(School of Materials and Chemical Engineering,Xuzhou Institute of Technology,Xuzhou 221018,China)

机构地区：[1]徐州工程学院化学化工学院,江苏徐州221018

出　　处：《徐州工程学院学报（自然科学版）》2023年第3期86-92,共7页Journal of Xuzhou Institute of Technology(Natural Sciences Edition)

基　　金：国家自然科学基金项目(20272095);江苏省自然科学基金项目(BK20171169)。

摘　　要：为了研究和开发醛糖还原酶抑制剂,从而设计高效抗糖尿病并发症的新药,采用拓扑理论计算了以112个羧酸类衍生物作为有潜力的ALR2抑制剂分子的电拓扑状态指数(Ei)和电性距离矢量(Mj),通过最佳变量子集回归的方法建立了这112个化合物生物活性(pIC50)的六元(M_(62),E_(13),M_(63),M_(14),E_(32),E_(42))QSAR模型,并以这6个参数作为输入层构建6∶9∶1的人工神经网络反向传播(BP)算法模型.该模型的相关系数由多元线性归回的0.859提升为0.979,预测能力由平均相对误差0.366降为0.119.通过对模型的6个参数进行分析,找出影响112个ALR2抑制剂分子生物活性的结构片段为—C—、—N—、>S<、—OH、—X,并以此设计了5个具有较高活性的分子.该研究为设计高效抗糖尿病并发症的新药提供了理论基础.In order to study and develop aldose reduction inhibitors and design new drugs with high efficiency against the complications of diabetes,the topological state index(Ei)and the electric distance vector(Mj)of 112 carboxylic acid derivatives as potential ALR2 inhibitors were calculated using topological theory,and the six component(M_(62),E_(13),M_(63),M_(14),E_(32),E_(42))QSAR model of the biological activity(pIC50)of these 112 compounds was established through the best variable subset regression method,These six parameters were used as the input neurons of artificial neural network,and a 6∶9∶1 network architecture was employed.The correlation coefficient of this model was increased from 0.859 in multivariate linear regression to 0.979,and the prediction ability was reduced from the average relative error of 0.366 to 0.119.By analyzing the six parameters of the model,the structural fragments affecting the biological activities of 112 ALR2 inhibitors were found to be—C—,—N—,>S<,—OH,—X,and five molecules with high activity were designed.This study provides a theoretical basis for the design of new drugs with high efficiency against the complications of diabetes.

关 键 词：醛糖还原酶抑制剂 羧酸衍生物 生物活性 QSAR 神经网络 

分 类 号：TQ426[化学工程]