Laboratory of Computational Biology, National Heart, Lung, and Blood Institute (NHLBI), National Institutes of Health (NIH), Bethesda, Maryland 20892, USA.
J Chem Phys. 2011 Apr 7;134(13):134108. doi: 10.1063/1.3574397.
This work derives a quantitative description of the conformational distribution in self-guided Langevin dynamics (SGLD) simulations. SGLD simulations employ guiding forces calculated from local average momentums to enhance low-frequency motion. This enhancement in low-frequency motion dramatically accelerates conformational search efficiency, but also induces certain perturbations in conformational distribution. Through the local averaging, we separate properties of molecular systems into low-frequency and high-frequency portions. The guiding force effect on the conformational distribution is quantitatively described using these low-frequency and high-frequency properties. This quantitative relation provides a way to convert between a canonical ensemble and a self-guided ensemble. Using example systems, we demonstrated how to utilize the relation to obtain canonical ensemble properties and conformational distributions from SGLD simulations. This development makes SGLD not only an efficient approach for conformational searching, but also an accurate means for conformational sampling.
这项工作推导出了自导朗之万动力学（SGLD）模拟中构象分布的定量描述。SGLD 模拟采用从局部平均动量计算得出的导向力来增强低频运动。这种低频运动的增强极大地加速了构象搜索效率，但也对构象分布产生了一定的干扰。通过局部平均，我们将分子系统的性质分为低频和高频两部分。使用这些低频和高频特性，我们定量描述了导向力对构象分布的影响。这种定量关系提供了在正则系综和自导系综之间转换的方法。通过示例系统，我们演示了如何利用这种关系从 SGLD 模拟中获得正则系综的性质和构象分布。这一发展使得 SGLD 不仅成为一种高效的构象搜索方法，而且成为一种准确的构象采样方法。
