药物分子胎盘屏障渗透的支持向量回归模型  

A prediction model of placenta barrier penetration of drug molecules based on support vector regression

作　　者：李伟[1] 周建平[1] 杨胜勇[1] 

机构地区：[1]四川大学华西医院生物治疗国家重点实验室,四川成都610041

出　　处：《化学研究与应用》2011年第7期872-876,共5页Chemical Research and Application

基　　金：国家自然科学基金项目(20872100)资助

摘　　要：药物分子透过胎盘屏障进入胎儿体内可能对胎儿造成不良影响,在临床及药物研发中评估药物分子的胎盘屏障渗透能力,对于保护受孕期的妇女以及胎儿,都有着十分重要的作用。本文使用了组合的遗传算法-共轭梯度-支持向量回归(GA-CG-SVR)方法,建立了一个药物分子胎盘屏障渗透的预测模型。对于所建模型,使用5重交叉验证和独立测试集方法进行了验证,预测值和实验值的相关系数分别达到0.82和0.77。我们也讨论了描述符选择和参数优化对SVR模型预测精度的影响,并将GA-CG-SVR模型与常用的多重线性回归(MLR)模型和K最邻近节点法(KNN)模型进行了比较,结果表明,GA-CG-SVR模型的预测精度明显优于其他方法建立的模型。Many drug molecules can transfer from the mother to her fetus across the placenta barrier,which may lead to toxic effects to the fetus.So it is important to evaluate the placenta barrier penetration of drug molecules for protecting the mother and the fetus.In this investigation,we adopt an support vector regression（SVR）combined with genetic algorithm（GA）and conjugate gradient（CG）method（GA-CG-SVR）to build a prediction model of placenta barrier penetration of drug molecules.Five-fold-cross-validation method and an independent evaluation set method were used to test the SVR model,which results show that the correlation coefficients of the predicted values by the developed model and experimental values of the training sets and the test sets are 0.82 and 0.77,respectively.We also discuss the influence of the descriptor selection and the SVM parameter setting on the prediction accuracy of SVR model.Finally,compared with models developed by multi-linear regression（MLR）and k-nearest neighbor（KNN）methods,the model of placenta barrier penetration of drug molecules built by GA-CG-SVR is obviously better in terms of the prediction accuracy.

关 键 词：胎盘屏障 支持向量回归 描述符选择 参数优化 

分 类 号：O641.121[理学—物理化学]