†Department of Applied Physics and ‡Department of Computational Science and Engineering, Nagoya University, Nagoya, Aichi 464-8603, Japan.
J Chem Inf Model. 2015 Jun 22;55(6):1108-19. doi: 10.1021/acs.jcim.5b00134. Epub 2015 Jun 9.
As the number of structurally resolved protein-ligand complexes increases, the ligand-binding pockets of many proteins have been found to accommodate multiple different compounds. Effective use of these structural data is important for developing virtual screening (VS) methods that identify bioactive compounds. Here, we introduce a VS method, VS-APPLE (Virtual Screening Algorithm using Promiscuous Protein-Ligand complExes), based on promiscuous protein-ligand binding structures. In VS-APPLE, multiple ligands bound to a pocket are combined into a query template for screening. Both the structural match between a test compound and the multiple-ligand template and the possible collisions between the test compound and the target protein are evaluated by an efficient geometric hashing method. The performance of VS-APPLE was examined on a filtered, clustered version of the Directory of Useful Decoys data set. In Area Under the Curve analyses of this data set, VS-APPLE outperformed several popular screening programs. Judging from the performance of VS-APPLE, the structural data of promiscuous protein-ligand bindings could be further analyzed and exploited for developing VS methods.
随着结构解析的蛋白-配体复合物数量的增加，许多蛋白质的配体结合口袋被发现能够容纳多种不同的化合物。有效利用这些结构数据对于开发虚拟筛选（VS）方法以识别生物活性化合物非常重要。在这里，我们介绍了一种基于混杂蛋白-配体结合结构的 VS 方法，VS-APPLE（使用混杂蛋白-配体复合物的虚拟筛选算法）。在 VS-APPLE 中，多个结合到口袋中的配体被组合成一个查询模板进行筛选。通过有效的几何哈希方法评估测试化合物与多配体模板之间的结构匹配以及测试化合物与靶蛋白之间可能发生的碰撞。在有用诱饵目录数据集的过滤、聚类版本上对 VS-APPLE 的性能进行了检查。在该数据集的曲线下面积分析中，VS-APPLE 优于几个流行的筛选程序。从 VS-APPLE 的性能来看，可以进一步分析和利用混杂蛋白-配体结合的结构数据来开发 VS 方法。
