Department of Chemistry, Stanford University , Stanford, California 94305, United States.
J Chem Theory Comput. 2015 Jun 9;11(6):2412-20. doi: 10.1021/acs.jctc.5b00031.
As simulators attempt to replicate the dynamics of large cellular components in silico, problems related to sampling slow, glassy degrees of freedom in molecular systems will be amplified manyfold. It is tempting to augment simulation techniques with external biases to overcome such barriers with ease; biased simulations, however, offer little utility unless equilibrium properties of interest (both kinetic and thermodynamic) can be recovered from the data generated. In this Article, we present a general scheme that harnesses the power of Markov state models (MSMs) to extract equilibrium kinetic properties from molecular dynamics trajectories collected on biased potential energy surfaces. We first validate our reweighting protocol on a simple two-well potential, and we proceed to test our method on potential-biased simulations of the Trp-cage miniprotein. In both cases, we find that equilibrium populations, time scales, and dynamical processes are reliably reproduced as compared to gold standard, unbiased data sets. We go on to discuss the limitations of our dynamical reweighting approach, and we suggest auspicious target systems for further application.
随着模拟器试图在计算机中复制大型细胞成分的动力学，与分子系统中缓慢的玻璃态自由度相关的问题将被放大许多倍。用外部偏置来增强模拟技术以轻松克服这些障碍是很诱人的；然而，偏置模拟除非可以从生成的数据中恢复出感兴趣的平衡性质（包括动力学和热力学），否则几乎没有用处。在本文中，我们提出了一种通用方案，利用马尔可夫状态模型（MSM）的强大功能从偏置势能表面上收集的分子动力学轨迹中提取平衡动力学性质。我们首先在一个简单的双势阱上验证了我们的重新加权方案，然后我们在 Trp-cage 小蛋白的势偏置模拟上测试了我们的方法。在这两种情况下，与黄金标准的无偏数据集相比，我们发现平衡种群、时间尺度和动力学过程都得到了可靠的再现。我们继续讨论我们的动力学重新加权方法的局限性，并提出了进一步应用的有希望的目标系统。
