MSU-DOE Plant Research Laboratory and Department of Biochemistry and Molecular Biology, Michigan State University, 612 Wilson Rd, East Lansing, Michigan 48824, United States.
J Chem Theory Comput. 2022 Oct 11;18(10):6161-6171. doi: 10.1021/acs.jctc.2c00327. Epub 2022 Sep 21.
Molecular simulation technologies have afforded researchers a unique look into the nanoscale interactions driving physical processes. However, a limitation for molecular dynamics (MD) simulations is that they must be performed on finite-sized systems in order to map onto computational resources. To minimize artifacts arising from finite-sized simulation systems, it is common practice for MD simulations to be performed with periodic boundary conditions (PBCs). However, in order to calculate specific physical properties, such as mean square displacements to calculate diffusion coefficients, continuous particle trajectories where the atomic movements are continuous and do not jump between cell faces are required. In these cases, modifying atomic coordinates through unwrapping schemes is an essential post-processing tool to remove these jumps. Here, two established trajectory unwrapping schemes are applied to 1 μs wrapped trajectories for a small water box and lysozyme in water. The existing schemes can result in spurious diffusion coefficients, long bonds within unwrapped molecules, and inconsistent atomic coordinates when coordinates are rewrapped after unwrapping. We determine that prior unwrapping schemes do not account for changing periodic box dimensions and introduce an additional correction term to the existing displacement unwrapping scheme to correct for these artifacts. We also demonstrate that the resulting algorithm is a hybrid between the existing heuristic and displacement unwrapping schemes. After treatment using this new unwrapping scheme, molecular geometries are correct even after long simulations. In anticipation for longer MD trajectories, we develop implementations for this new scheme in multiple PBC handling tools.
分子模拟技术为研究人员提供了一种独特的方法，可以深入了解推动物理过程的纳米尺度相互作用。然而，分子动力学（MD）模拟的一个限制是，为了映射到计算资源，它们必须在有限大小的系统上进行。为了最小化来自有限大小模拟系统的伪影，MD 模拟通常采用周期性边界条件（PBC）。然而，为了计算特定的物理性质，例如均方位移以计算扩散系数，需要连续的粒子轨迹，其中原子运动是连续的，不会在细胞面之间跳跃。在这些情况下，通过展开方案修改原子坐标是去除这些跳跃的必要后处理工具。在这里，两种成熟的轨迹展开方案被应用于一个小的水盒和水中溶菌酶的 1 μs 包裹轨迹。现有的方案可能会导致虚假的扩散系数、展开分子中的长键以及在展开后重新包裹坐标时坐标不一致。我们确定现有的展开方案没有考虑到周期性盒子尺寸的变化，并引入了一个附加的校正项到现有的位移展开方案中，以纠正这些伪影。我们还证明，该算法是现有启发式和位移展开方案之间的混合算法。在使用这种新的展开方案进行处理后，即使在长时间的模拟后，分子几何形状也是正确的。为了应对更长的 MD 轨迹，我们在多个 PBC 处理工具中为该新方案开发了实现。
