论文部分内容阅读
通过构建基于分子属性的分类树模型以鉴别化合物的生物药剂分类系统(biopharmaceutics classification system,BCS)的穿透性分类。将从不同文献采集的Caco-2穿透性数据构成训练集,建立分类树模型,并应用此模型对外部测试集——美国食品药品监督管理局BCS的标准化合物进行分类测试。由此建立的鉴别化合物的BCS穿透性分类的规则为:如果氢键供体原子数量<4、正性范德华极性表面积和<40.71并且溶解能>-33.89,那么该化合物为高穿透性,否则为低穿透性。本分类结构属性关系模型的105个化合物的训练集和17个化合物的外部测试集的识别正确率分别91.43%和82.35%。本模型成功应用于鉴定BCS标准化合物高低穿透性分类药物的分子属性,为药物穿透性的识别提供了简便、有效的分类方法。
The classification of penetrability of the compound biopharmaceutics classification system (BCS) was identified by constructing a classification tree model based on molecular properties. Caco-2 penetrating data collected from different sources were used to construct a training set, and a classification tree model was established. The model was used to test the external test set - the standard compound of the US Food and Drug Administration’s BCS. The established rule for BCS penetrance classification of identified compounds is that the compound is highly penetrating if the number of hydrogen bond donor atoms is <4 and the positive van der Waals’ polar surface area is <40.71 and the dissolution energy is> -33.89 , Otherwise low penetration. The recognition accuracy of the training set of 105 compounds and the external test set of 17 compounds in the classification structure-attribute relational model were 91.43% and 82.35% respectively. The model has been successfully applied to identify the molecular properties of BCS standard compound penetrating drugs, which provides a simple and effective classification method for the drug penetrant identification.