【作者】 张凯；

【导师】 向大雄；

【作者基本信息】 中南大学 ， 药学， 2014， 硕士

【摘要】 一、研究背景及目的本课题采用MOE2011,从期刊文献数据库,化学信息学数据库等收集MDCK细胞渗透性数据,计算并筛选分子描述符,进行MDCK细胞渗透性模型的建立并评价。本课题旨在建立一个稳健、可靠的MDCK细胞渗透性模型,为药物早期发现过程中筛选提供一个MDCK细胞渗透性数据评价的有力工具。二、方法与结果1.MDCK细胞渗透性性模型的建立本章通过网络进行MDCK细胞渗透性数据收集,并进行格式转换,并输入。采用MOE2011进行分子能量优化,分子描述符(包括2D分子描述符和3D分子描述符)的计算,并随机划分为训练集和测试集。去除重复,或相关性较大的描述符,或为常量的分子描述符,后采用GA-PLA进行分子描述符的筛选,最终获得10个分子描述符。采用MLR进行线性模型的建立。模型的均方根误差(RMSE)为0.43207,相关系数R2为0.86628。2.MDCK细胞渗透性模型的评价本章采用MDE2011AutoQS/AR模块,用LOO验证法,5折交叉验证法,Y随机化法,测试集验证法,模型应用域等进行MDCK细胞渗透性模型的评价。LOO RMSE为0.52019,LOO R2为0.80768。5折交叉验证法的RMSE为0.52019。Y随机化法的R2与模型的R2差异显著。测试集验证法的相关系数q2为0.77115。三、结论本课题建立了MDCK细胞渗透性模型：-logPapp=-2.81270+1.03126*PC+-0.01293*PEOE VSA+1+0.41244*dipoleX+5.31619*glob-0.10999*opr_brigid+2.45322*vsurf_CW1+0.02388*vsurf_D8+1.05636*vsurf-EDminl-0.04246*vsurf_Wp4+0.02238*SMR VSA4.LOO验证法,5折交叉验证法,Y随机化法,测试集验证等验证后,该模型仍稳定可靠。说明该模型稳健,鲁棒性强。本课题为药物早期发现中化合物MDCK细胞渗透性预测提供参考。更多还原

【Abstract】 BACKGROUND AND OBJECTIVIESIn this thesis we collected MDCK cell permeability data from journal database and chemoinfromatics database using MOE2011, caculated molecular descriptors, established and evaluated a MDCK cell permeability model. This paper aims to establish a roboust and precise QSAR model of MDCK cell permeability, offer a tool to screen compounds MDCK cell permeability in silico in early drug discovery and develop period. METHODS AND RESULTS1. Estalishment of MDCK cell permeability modelIn this chapter MDCK cell permeability data were collected,transformed and input.Using MOE2011chemical structure was energy minimized. Caculate the molecular descriptors (including2D molecular and3D molecular descriptors), and spilt into training and test set randomly. Remove the duplication, involve descriptors, and nearly constant descriptors, then screen the variability using GA-PLA, finally remain10descriptors. Create a liner model using MLR. The root mean square error (RMSE) is0.43207,and the correlation coefficient (R2) is0.86628.2. Evaluation of MDCK cell permeabiltity modelIn this chapter the MDCK cell permeability model was evaluated using LOO,5fold validation, Y randomize, test set validation and application domain in MOE2011AutoQSAR moduel. The outcome is below:LOO RMSE=0.52019; LOO R2=0.80768;5fold RMSE=0.52019; Z-SCORE VS RANDOM=9.64223; q2=0.77115.CONCLUSIONSThis theis estabish a MDCK cell permeability model:-logPapp=-2.81270+1.03126*PC+-0.01293*PEOE_VSA+1+0.41244*dipoleX+5.31619 *glob-0.10999*opr_brigid+2.45322*vsurf_CW1+0.02388*vsurf_D8+1.05636*vsurfJEDminl-0.04246*vsurf_Wp4+0.02238*SMR_VSA4. After LOO validation,5fold validation, Y randomize validation and application domain, the model acts well. So the model is robust. This paper offers an advise to compounds to screen MDCK cell permeability in drug discovery and develop.更多还原