Centro de Investigaciones Biológicas, CSIC, Ramiro de Maeztu, 9. 28040 Madrid, Spain.
Bioinformatics. 2009 Oct 1;25(19):2544-51. doi: 10.1093/bioinformatics/btp447. Epub 2009 Jul 20.
Prediction of protein-protein complexes from the coordinates of their unbound components usually starts by generating many potential predictions from a rigid-body 6D search followed by a second stage that aims to refine such predictions. Here, we present and evaluate a new method to effectively address the complexity and sampling requirements of the initial exhaustive search. In this approach we combine the projection of the interaction terms into 3D grid-based potentials with the efficiency of spherical harmonics approximations to accelerate the search. The binding energy upon complex formation is approximated as a correlation function composed of van der Waals, electrostatics and desolvation potential terms. The interaction-energy minima are identified by a novel, fast and exhaustive rotational docking search combined with a simple translational scanning. Results obtained on standard protein-protein benchmarks demonstrate its general applicability and robustness. The accuracy is comparable to that of existing state-of-the-art initial exhaustive rigid-body docking tools, but achieving superior efficiency. Moreover, a parallel version of the method performs the docking search in just a few minutes, opening new application opportunities in the current 'omics' world.
从其未结合成分的坐标预测蛋白质-蛋白质复合物通常首先从刚体 6D 搜索生成许多潜在的预测开始,然后进入第二阶段,旨在改进此类预测。在这里,我们提出并评估了一种新的方法,以有效地解决初始穷举搜索的复杂性和采样要求。在这种方法中,我们将相互作用项的投影与基于 3D 网格的势的效率结合起来,使用球谐逼近来加速搜索。复合物形成时的结合能被近似为由范德华、静电和去溶剂化势项组成的相关函数。通过一种新颖的、快速的和全面的旋转对接搜索与简单的平移扫描相结合,确定了相互作用能的最小值。在标准蛋白质-蛋白质基准测试中获得的结果证明了其通用性和鲁棒性。其准确性可与现有的最先进的初始刚体对接工具相媲美,但效率更高。此外,该方法的并行版本仅需几分钟即可执行对接搜索,为当前的“组学”世界开辟了新的应用机会。