摘要
以24个作用于人体外围血单核细胞药理模型的白介素-1β转化酶抑制剂作为研究对象,计算了其表征分子的拓扑、电子、几何结构等物理化学性质的1209个分子描述符,用CfsSubsetEval评价方法和BestFirst-D1-N5搜索方法筛选描述符,用Kennard-Stone方法选择训练集和测试集。分别采用支持向量机、决策树、贝叶斯网络、人工神经网络等机器学习方法建立分类预测模型并使用Catalyst/HipHop系统建立药效团模型。结果表明支持向量机优于其他分类模型,正、负样本的预测正确率均达到100%。最优药效团模型具有5个特征:2个疏水基团、2个脂性氢键受体、1个氢键给体;以此药效团进行中药数据库筛选得到384个候选白介素-1β转化酶抑制剂。利用支持向量机建立的分类预测模型对候选化合物的活性进行了预测,其中高活性化合物占96.6%,表明白介素-1β转化酶抑制剂药效团模型较准确地反映了高活性化合物的公共特征。该模型的建立有助于从中草药筛选新型白介素-1β转化酶抑制剂。
1209 molecular descriptors, including topological descriptors, electronic descriptors, and geometric descriptors, were calculated to characterize the physicochemical properties for 24 ICE inhibitors from human peripheral blood monocytes. CfsSubsetEval evaluation method and BestFirst-D1-N5 search method were applied to the variable selection. The Kennard-Stone method was adopted to select the training set and the testing set. Machine learning methods, including the Support Vector Machine （SVM）, Decision Tree, Bayesian Network, and Artificial Neural Network, were used to develop the classification models. Meanwhile, three-dimensional pharmacophore models were generated by program Catalyst/ HipHop. It was shown that the SVM outperformed other classification models and the prediction accuracy of positive and negative samples reached 100%. The best pharmacophore consisted of five features： two hydrophobic （H）, two hydrogenbond acceptor lipid （Ali）, and one hydrogen-bond donor （D）. 384 candidate ICE inhibitors were selected from the Traditional Chinese Medicine Database by the pharmacophore. On the other hand, using the SVM classification model to predict the activity of candidate compounds, highly active compounds accounted for 96.6%. The result suggested that the ICE inhibitors pharmacophore model accurately reflected the characteristics of highly active compounds. The pharmacophore model may contribute to screening of new ICE inhibitors from the Traditional Chinese Medicine.
出处
《世界科学技术-中医药现代化》
2009年第6期783-788,共6页
Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology
基金
科技部国家"973"计划(2005CB523401):组分配伍与饮片配伍的相关性研究
负责人:郑虎占
国家"973"计划(2006CB504703):寒热药性的内在规律及共同属性研究
负责人:乔延江
中医药行业科研专项(200707010):针对病毒性疾病的中药活性发现关键技术研究
负责人:朱晓新
