Department of Bioengineering, University of California, Berkeley, California, USA.
J Comput Chem. 2010 Feb;31(3):620-7. doi: 10.1002/jcc.21355.
We have devised a systematic approach to converge a replica exchange molecular dynamics simulation by dividing the full temperature range into a series of higher temperature reservoirs and a finite number of lower temperature subreplicas. A defined highest temperature reservoir of equilibrium conformations is used to help converge a lower but still hot temperature subreplica, which in turn serves as the high-temperature reservoir for the next set of lower temperature subreplicas. The process is continued until an optimal temperature reservoir is reached to converge the simulation at the target temperature. This gradual convergence of subreplicas allows for better and faster convergence at the temperature of interest and all intermediate temperatures for thermodynamic analysis, as well as optimizing the use of multiple processors. We illustrate the overall effectiveness of our multiple reservoir replica exchange strategy by comparing sampling and computational efficiency with respect to replica exchange, as well as comparing methods when converging the structural ensemble of the disordered Abeta(21-30) peptide simulated with explicit water by comparing calculated Rotating Overhauser Effect Spectroscopy intensities to experimentally measured values.
我们设计了一种系统的方法，通过将全温度范围划分为一系列高温储层和有限数量的低温子副本，来收敛 replica 交换分子动力学模拟。使用定义的最高温度平衡构象储层来帮助收敛较低但仍然很热的温度子副本，该副本反过来又成为下一组较低温度子副本的高温储层。该过程持续进行，直到达到最佳温度储层，以在目标温度下收敛模拟。这种子副本的逐渐收敛允许在感兴趣的温度和所有中间温度下进行更好和更快的收敛，用于热力学分析，以及优化多个处理器的使用。我们通过 replica 交换比较采样和计算效率，以及在通过比较计算的旋转过氧化物效应光谱强度与实验测量值来模拟无序 Abeta(21-30)肽的结构集合时比较方法，来展示我们的多储层 replica 交换策略的整体有效性。
