Department of Applied Mathematics and Statistics, Stony Brook University, 11794 Stony Brook, NY, USA; Laufer Center for Physical and Quantitative Biology, Stony Brook University, 11794 Stony Brook, NY, USA; Institute for Advanced Computational Sciences, Stony Brook University, 11794 Stony Brook, NY, USA.
Moscow Institute of Physics and Technology, 141701 Moscow, Russia.
J Mol Biol. 2018 Jul 20;430(15):2249-2255. doi: 10.1016/j.jmb.2018.03.010. Epub 2018 Apr 5.
We have recently demonstrated that incorporation of small-angle X-ray scattering (SAXS)-based filtering in our heavily used docking server ClusPro improves docking results. However, the filtering step is time consuming, since ≈10 conformations have to be sequentially processed. At the same time, we have demonstrated the possibility of ultra-fast systematic energy evaluation for all rigid body orientations of two proteins, by sampling using Fast Manifold Fourier Transform (FMFT), if energies are represented as a combination of convolution-like expressions. Here we present a novel FMFT-based algorithm FMFT-SAXS for massive SAXS computation on multiple conformations of a protein complex. This algorithm exploits the convolutional form of SAXS calculation function. FMFT-SAXS allows computation of SAXS profiles for millions of conformations in a matter of minutes, providing an opportunity to explore the whole conformational space of two interacting proteins. We demonstrate the application of the new FMFT-SAXS approach to significantly speed up SAXS filtering step in our current docking protocol (1 to 2 orders of magnitude faster, running in several minutes on a modern 16-core CPU) without loss of accuracy. This is demonstrated on the benchmark set as well as on the experimental data. The new approach is available as a part of ClusPro server (https://beta.cluspro.org) and as an open source C library (https://bitbucket.org/abc-group/libfmftsaxs).
我们最近证明，在我们广泛使用的对接服务器 ClusPro 中加入小角 X 射线散射（SAXS）过滤技术可以提高对接结果。然而，过滤步骤非常耗时，因为需要依次处理 ≈10 个构象。同时，如果能量可以表示为卷积表达式的组合，我们已经证明了使用快速流形傅里叶变换（FMFT）对两个蛋白质的所有刚体取向进行超快速系统能量评估的可能性。在这里，我们提出了一种新的基于 FMFT 的算法 FMFT-SAXS，用于对蛋白质复合物的多个构象进行大规模的 SAXS 计算。该算法利用了 SAXS 计算函数的卷积形式。FMFT-SAXS 允许在几分钟内计算数百万个构象的 SAXS 谱图，从而有机会探索两个相互作用蛋白质的整个构象空间。我们展示了新的 FMFT-SAXS 方法在我们当前对接方案中的 SAXS 过滤步骤中的应用，速度可以提高 1 到 2 个数量级（在现代 16 核 CPU 上运行几分钟），而不会降低准确性。这在基准集和实验数据上都得到了验证。该新方法可作为 ClusPro 服务器（https://beta.cluspro.org）的一部分以及开源 C 库（https://bitbucket.org/abc-group/libfmftsaxs）使用。
