Ban Tomohiro, Ohue Masahito, Akiyama Yutaka
School of Computing, Tokyo Institute of Technology, 2-12-1 W8-76 Ookayama, Meguro-ku, Tokyo 152-8550, Japan; Education Academy of Computational Life Sciences, Tokyo Institute of Technology, 2-12-1 W8-93 Ookayama, Meguro-ku, Tokyo 152-8550, Japan.
School of Computing, Tokyo Institute of Technology, 2-12-1 W8-76 Ookayama, Meguro-ku, Tokyo 152-8550, Japan; Advanced Computational Drug Discovery Unit, Institute of Innovative Research, Tokyo Institute of Technology, 4259 Nagatsutacho, Midori-ku, Yokohama, Kanagawa 226-8501, Japan.
Comput Biol Chem. 2018 Apr;73:139-146. doi: 10.1016/j.compbiolchem.2018.02.008. Epub 2018 Feb 15.
The identification of comprehensive drug-target interactions is important in drug discovery. Although numerous computational methods have been developed over the years, a gold standard technique has not been established. Computational ligand docking and structure-based drug design allow researchers to predict the binding affinity between a compound and a target protein, and thus, they are often used to virtually screen compound libraries. In addition, docking techniques have also been applied to the virtual screening of target proteins (inverse docking) to predict target proteins of a drug candidate. Nevertheless, a more accurate docking method is currently required. In this study, we proposed a method in which a predicted ligand-binding site is covered by multiple grids, termed multiple grid arrangement. Notably, multiple grid arrangement facilitates the conformational search for a grid-based ligand docking software and can be applied to the state-of-the-art commercial docking software Glide (Schrödinger, LLC). We validated the proposed method by re-docking with the Astex diverse benchmark dataset and blind binding site situations, which improved the correct prediction rate of the top scoring docking pose from 27.1% to 34.1%; however, only a slight improvement in target prediction accuracy was observed with inverse docking scenarios. These findings highlight the limitations and challenges of current scoring functions and the need for more accurate docking methods. The proposed multiple grid arrangement method was implemented in Glide by modifying a cross-docking script for Glide, xglide.py. The script of our method is freely available online at http://www.bi.cs.titech.ac.jp/mga_glide/.
全面识别药物-靶点相互作用在药物发现中至关重要。尽管多年来已开发出众多计算方法，但尚未建立起金标准技术。计算配体对接和基于结构的药物设计使研究人员能够预测化合物与靶蛋白之间的结合亲和力，因此，它们常被用于虚拟筛选化合物库。此外，对接技术也已应用于靶蛋白的虚拟筛选（反向对接），以预测候选药物的靶蛋白。然而，目前需要一种更精确的对接方法。在本研究中，我们提出了一种方法，其中预测的配体结合位点由多个网格覆盖，称为多网格排列。值得注意的是，多网格排列有助于基于网格的配体对接软件进行构象搜索，并且可以应用于最先进的商业对接软件Glide（Schrödinger有限责任公司）。我们通过使用阿斯泰克斯多样化基准数据集重新对接和盲结合位点情况验证了所提出的方法，这将得分最高的对接姿势的正确预测率从27.1%提高到了34.1%；然而，在反向对接情况下，仅观察到靶标预测准确性有轻微提高。这些发现突出了当前评分函数的局限性和挑战以及对更精确对接方法的需求。所提出的多网格排列方法通过修改Glide的交叉对接脚本xglide.py在Glide中实现。我们方法的脚本可在http://www.bi.cs.titech.ac.jp/mga_glide/在线免费获取。
