Department of Biological Science and Engineering, School of Chemistry and Biological Engineering , University of Science and Technology Beijing , Beijing 100083 , China.
Beijing Key Lab of Bioprocess, College of Life Science and Technology , Beijing University of Chemical Technology , Box 53, Beijing 100029 , China.
J Chem Inf Model. 2018 May 29;58(5):1037-1052. doi: 10.1021/acs.jcim.8b00026. Epub 2018 Apr 18.
Thermodynamic and kinetic properties are of critical importance for the applicability of computational models to biomolecules such as proteins. Here we present an extensive evaluation of the Amber ff99SB-ILDN force field for modeling of hydration and diffusion of amino acids with three-site (SPC, SPC/E, SPC/E, and TIP3P), four-site (TIP4P, TIP4P-Ew, and TIP4P/2005), and five-site (TIP5P and TIP5P-Ew) water models. Hydration free energies (HFEs) of neutral amino acid side chain analogues have little dependence on the water model, with a root-mean-square error (RMSE) of ∼1 kcal/mol from experimental observations. On the basis of the number of interacting sites in the water model, HFEs of charged side chains can be putatively classified into three groups, of which the group of three-site models lies between those of four- and five-site water models; for each group, the water model dependence is greatly eliminated when the solvent Galvani potential is considered. Some discrepancies in the location of the first hydration peak ( R) in the ion-water radial distribution function between experimental and calculated observations were detected, such as a systematic underestimation of the acetate (Asp side chain) ion. The RMSE of calculated diffusion coefficients of amino acids from experiment increases linearly with the increasing diffusion coefficients of the solvent water models at infinite dilution. TIP3P has the fastest diffusivity, in line with literature findings, while the "FB" and "OPC" water model families as well as TIP4P/2005 perform well, within a relative error of 5%, and TIP4P/2005 yields the most accurate estimate for the water diffusion coefficient. All of the tested water models overestimate amino acid diffusion coefficients by approximately 40% (TIP4P/2005) to 200% (TIP3P). Scaling of protein-water interactions with TIP4P/2005 in the Amber ff99SBws and ff03ws force fields leads to more negative HFEs but has little influence on the diffusion of amino acids. The most recent FF/water combinations of ff14SB/OPC3, ff15ipq/SPC/E, and fb15/TIP3P-FB do not show obvious improvements in accuracy for the tested quantities. These findings here establish a benchmark that may aid in the development and improvement of classical force fields to accurately model protein dynamics and thermodynamics.
热力学和动力学性质对于计算模型在生物分子（如蛋白质）中的应用至关重要。在这里，我们广泛评估了 Amber ff99SB-ILDN 力场，以模拟具有三站点（SPC、SPC/E、SPC/E 和 TIP3P）、四站点（TIP4P、TIP4P-Ew 和 TIP4P/2005）和五站点（TIP5P 和 TIP5P-Ew）的水模型的氨基酸的水合和扩散。中性氨基酸侧链类似物的水合自由能（HFEs）几乎不受水模型的影响，与实验观察值的均方根误差（RMSE）约为 1 kcal/mol。根据水模型中相互作用的站点数量，可以将带电荷侧链的 HFEs 假定分为三组，其中三站点模型组介于四站点和五站点水模型组之间；对于每组，当考虑溶剂 Galvani 势时，水模型的依赖性大大消除。在实验和计算观测到的离子-水径向分布函数中，第一水合峰（R）的位置存在一些差异，例如乙酸盐（Asp 侧链）离子的系统低估。从实验计算得到的氨基酸扩散系数的 RMSE 与溶剂水模型的无限稀释扩散系数呈线性增加。TIP3P 的扩散速度最快，与文献发现一致，而“FB”和“OPC”水模型家族以及 TIP4P/2005 表现良好，相对误差在 5%以内，TIP4P/2005 对水扩散系数的估计最为准确。所有测试的水模型都将氨基酸扩散系数高估了约 40%（TIP4P/2005）到 200%（TIP3P）。在 Amber ff99SBws 和 ff03ws 力场中用 TIP4P/2005 对蛋白质-水相互作用进行缩放导致 HFEs 更负，但对氨基酸的扩散影响很小。最新的 ff14SB/OPC3、ff15ipq/SPC/E 和 fb15/TIP3P-FB 的 FF/water 组合在测试量的准确性方面没有明显提高。这些发现为开发和改进经典力场以准确模拟蛋白质动力学和热力学提供了基准。
