Efficient Determination of Free Energy Landscapes in Multiple Dimensions from Biased Umbrella Sampling Simulations Using Linear Regression.
Efficient Determination of Free Energy Landscapes in Multiple Dimensions from Biased Umbrella Sampling Simulations Using Linear Regression.
复制标题
DOI：
10.1021/ct501130r
复制
发表时间：
2015-08-11
影响因子：
5.5
通讯作者：
Roux B
中科院分区：
文献类型：
--
作者：
Meng Y;Roux B
加权直方图分析方法 (WHAM) 是一种标准协议，用于对有偏伞采样模拟的信息进行后处理，以构建相对于一组阶次参数的平均力的潜力。借助 WHAM 方程，通过迭代过程满足自洽条件来确定无偏态密度。虽然该方法在阶数参数数量较少时非常有效，但在更高维度时其计算成本快速增长。在这里，我们提出了一种简单而有效的替代策略，它避免了迭代求解自洽 WHAM 方程。利用有效的多元线性回归框架来链接各个伞窗的有偏概率密度，并在序参数空间中产生无偏的全局自由能景观。通过实际例子证明，可以以一小部分成本生成准确度与 WHAM 相当的自由能源景观。
The weighted histogram analysis method (WHAM) is a standard protocol for postprocessing the information from biased umbrella sampling simulations to construct the potential of mean force with respect to a set of order parameters. By virtue of the WHAM equations, the unbiased density of state is determined by satisfying a self-consistent condition through an iterative procedure. While the method works very effectively when the number of order parameters is small, its computational cost grows rapidly in higher dimension. Here, we present a simple and efficient alternative strategy, which avoids solving the self-consistent WHAM equations iteratively. An efficient multivariate linear regression framework is utilized to link the biased probability densities of individual umbrella windows and yield an unbiased global free energy landscape in the space of order parameters. It is demonstrated with practical examples that free energy landscapes that are comparable in accuracy to WHAM can be generated at a small fraction of the cost.