Söderholm Annu A, Lehtovuori Pekka T, Nyrönen Tommi H
CSC, Scientific Computing Ltd., P.O. Box 405, FI-02101 Espoo, Finland.
J Med Chem. 2006 Jul 13;49(14):4261-8. doi: 10.1021/jm060234e.
We report a docking and comparative molecular similarity indices analysis (CoMSIA) study of progesterone receptor (PR) ligands with an emphasis on nonsteroids including tanaproget. The ligand alignment generation, a critical part of model building, comprised two stages. First, thorough conformational sampling of docking poses within the PR binding pocket was made with the program GOLD. Second, a strategy to select representative poses for CoMSIA was developed utilizing the FlexX scoring function. After manual replacement of five poses where this approach had problems, a significant correlation (r(2) = 0.878) between the experimental affinities and electrostatic, hydrophobic, and hydrogen bond donor properties of the aligned ligands was found. Extensive model validation was made using random-group cross-validations, external test set predictions (r(pred)(2) = 0.833), and consistency check between the CoMSIA model and the PR binding site structure. Robustness, predictive ability, and automated alignment generation make the model a potential tool for virtual screening.
我们报告了一项关于孕激素受体（PR）配体的对接和比较分子相似性指数分析（CoMSIA）研究，重点是包括他那孕酮在内的非甾体类化合物。配体比对生成是模型构建的关键部分，包括两个阶段。首先，使用GOLD程序对PR结合口袋内的对接构象进行全面的构象采样。其次，利用FlexX评分函数开发了一种为CoMSIA选择代表性构象的策略。在手动替换该方法存在问题的五个构象后，发现对齐配体的实验亲和力与静电、疏水和氢键供体性质之间存在显著相关性（r(2) = 0.878）。使用随机分组交叉验证、外部测试集预测（r(pred)(2) = 0.833）以及CoMSIA模型与PR结合位点结构之间的一致性检查进行了广泛的模型验证。稳健性、预测能力和自动比对生成使得该模型成为虚拟筛选的潜在工具。
