Departamento de Farmacología, Universidad de Alcalá, E-28871 Alcalá de Henares, Madrid, Spain.
J Chem Inf Model. 2012 Aug 27;52(8):2300-9. doi: 10.1021/ci300194a. Epub 2012 Jul 19.
An ultrafast docking and virtual screening program, CRDOCK, is presented that contains (1) a search engine that can use a variety of sampling methods and an initial energy evaluation function, (2) several energy minimization algorithms for fine tuning the binding poses, and (3) different scoring functions. This modularity ensures the easy configuration of custom-made protocols that can be optimized depending on the problem in hand. CRDOCK employs a precomputed library of ligand conformations that are initially generated from one-dimensional SMILES strings. Testing CRDOCK on two widely used benchmarks, the ASTEX diverse set and the Directory of Useful Decoys, yielded a success rate of ~75% in pose prediction and an average AUC of 0.66. A typical ligand can be docked, on average, in just ~13 s. Extension to a representative group of pharmacologically relevant G protein-coupled receptors that have been recently cocrystallized with some selective ligands allowed us to demonstrate the utility of this tool and also highlight some current limitations. CRDOCK is now included within VSDMIP, our integrated platform for drug discovery.
我们介绍了一个超快对接和虚拟筛选程序 CRDOCK，它包含（1）一个可以使用多种采样方法和初始能量评估函数的搜索引擎，（2）几种用于微调结合构象的能量最小化算法，以及（3）不同的评分函数。这种模块化确保了可以根据手头的问题轻松配置定制协议，并且可以对其进行优化。CRDOCK 使用预先计算的配体构象库，这些构象最初是从一维 SMILES 字符串生成的。在两个广泛使用的基准测试（ASTEX 多样性集和有用诱饵目录）上测试 CRDOCK，在构象预测方面的成功率约为 75%，平均 AUC 为 0.66。一个典型的配体平均只需 13 秒即可对接。扩展到一组最近与一些选择性配体共结晶的具有药理学相关性的 G 蛋白偶联受体，使我们能够展示该工具的实用性，并突出一些当前的局限性。CRDOCK 现在包含在我们的药物发现集成平台 VSDMIP 中。
