Department of Chemistry and Chemical Biology, Rutgers University, 174 Frelinghueysen Road, Piscataway, New Jersey 08854-8066, USA.
Epic Systems, 1979 Milky Way, Verona, Wisconsin 53593, USA.
J Chem Phys. 2017 Oct 28;147(16):161730. doi: 10.1063/1.4985866.
The ff15ipq protein force field is a fixed charge model built by automated tools based on the two charge sets of the implicitly polarized charge method: one set (appropriate for vacuum) for deriving bonded parameters and the other (appropriate for aqueous solution) for running simulations. The duality is intended to treat water-induced electronic polarization with an understanding that fitting data for bonded parameters will come from quantum mechanical calculations in the gas phase. In this study, we compare ff15ipq to two alternatives produced with the same fitting software and a further expanded data set but following more conventional methods for tailoring bonded parameters (harmonic angle terms and torsion potentials) to the charge model. First, ff15ipq-Qsolv derives bonded parameters in the context of the ff15ipq solution phase charge set. Second, ff15ipq-Vac takes ff15ipq's bonded parameters and runs simulations with the vacuum phase charge set used to derive those parameters. The IPolQ charge model and associated protocol for deriving bonded parameters are shown to be an incremental improvement over protocols that do not account for the material phases of each source of their fitting data. Both force fields incorporating the polarized charge set depict stable globular proteins and have varying degrees of success modeling the metastability of short (5-19 residues) peptides. In this particular case, ff15ipq-Qsolv increases stability in a number of α-helices, correctly obtaining 70% helical character in the K19 system at 275 K and showing appropriately diminishing content up to 325 K, but overestimating the helical fraction of AAQAA by 50% or more, forming long-lived α-helices in simulations of a β-hairpin, and increasing the likelihood that the disordered p53 N-terminal peptide will also form a helix. This may indicate a systematic bias imparted by the ff15ipq-Qsolv parameter development strategy, which has the hallmarks of strategies used to develop other popular force fields, and may explain some of the need for manual corrections in this force fields' evolution. In contrast, ff15ipq-Vac incorrectly depicts globular protein unfolding in numerous systems tested, including Trp cage, villin, lysozyme, and GB3, and does not perform any better than ff15ipq or ff15ipq-Qsolv in tests on short peptides. We analyze the free energy surfaces of individual amino acid dipeptides and the electrostatic potential energy surfaces of each charge model to explain the differences.
ff15ipq 蛋白力场是一种基于隐极化电荷法的两个电荷集构建的固定电荷模型：一个电荷集（适用于真空）用于推导键参数，另一个电荷集（适用于水溶液）用于进行模拟。这种对偶性旨在用电子极化的概念来处理水诱导的电子极化，即理解键参数的拟合数据将来自气相的量子力学计算。在这项研究中，我们将 ff15ipq 与另外两种方法进行了比较，这两种方法使用相同的拟合软件和进一步扩展的数据集，但对键参数（谐波角项和扭转势）的拟合方法更加传统。首先，ff15ipq-Qsolv 在 ff15ipq 溶液相电荷集的背景下推导键参数。其次，ff15ipq-Vac 采用 ff15ipq 的键参数，并使用用于推导这些参数的真空相电荷集进行模拟。IPolQ 电荷模型和相关的键参数推导协议被证明是一种改进，它比不考虑拟合数据来源的材料相的协议更好。这两种力场都包含极化电荷集，在模拟稳定的球状蛋白方面都有不同程度的成功，在模拟短肽（5-19 个残基）的亚稳态方面也有不同程度的成功。在这种特殊情况下，ff15ipq-Qsolv 增加了许多 α-螺旋的稳定性，在 275 K 时正确获得了 K19 系统 70%的螺旋特征，并在 325 K 时显示出适当减少的含量，但对 AAQAA 的螺旋分数高估了 50%或更多，在 β-发夹模拟中形成了长寿命的 α-螺旋，并增加了无序的 p53 N 端肽也形成螺旋的可能性。这可能表明 ff15ipq-Qsolv 参数开发策略所带来的系统偏差，该策略具有开发其他流行力场策略的特征，这可能解释了在该力场的发展过程中需要手动修正的一些原因。相比之下，ff15ipq-Vac 在测试的多个系统中错误地描绘了球状蛋白的展开，包括色氨酸笼、微管蛋白、溶菌酶和 GB3，在短肽测试中也不比 ff15ipq 或 ff15ipq-Qsolv 表现得更好。我们分析了单个氨基酸二肽的自由能表面和每个电荷模型的静电势能表面，以解释这些差异。
