Scalable Indirect Free Energy Method Applied to Divalent Cation-Metalloprotein Binding.
Scalable Indirect Free Energy Method Applied to Divalent Cation-Metalloprotein Binding.
复制标题
DOI：
10.1021/acs.jctc.9b00147
复制
发表时间：
2019-06
影响因子：
5.5
通讯作者：
Jacob M. Litman;Andrew C Thiel;M. Schnieders
中科院分区：
文献类型：
--
作者：
Jacob M. Litman;Andrew C Thiel;M. Schnieders
许多生物过程是基于高电荷分子之间的分子识别，如核酸，无机离子，带电荷的氨基酸等，对于这种情况下，它已被证明，分子模拟与固定的部分电荷往往无法达到实验的准确性。虽然引入了更先进的静电模型（如多极、互极化等），可以显著提高模拟精度，但它使计算费用增加了5-20倍。间接自由能（IFE）方法可以通过在固定电荷分辨率下对中间状态进行建模来减轻这种成本。例如，有效的“参考”模型，如成对的琥珀，CHARMM，或OPLS-AA力场，可以用来推导出初始估计，然后通过热力学修正到更先进的“目标”潜力，如极化AMOEBA模型。不幸的是，所有目前描述的IFE方法遇到的困难，重新加权之间的分辨率超过150个原子，由于广泛的缩放的热力学修正的幅度和它们的统计不确定性。我们提出了一种方法，称为“同步bookending”（SB），这是从根本上不同于现有的IFE方法的基础上可调采样近似，它允许缩放到数千个原子。SB被证明在Mg 2 +/Ca 2+的相对结合亲和力的一组金属蛋白与多达2972个原子，发现直接AMOEBA的结果之间没有统计学上的显着差异，从修正琥珀色AMOEBA。在重新加权期间改变数千个原子的分辨率的能力表明，该方法可能在未来适用于蛋白质-蛋白质结合亲和力或核酸热力学。
Many biological processes are based on molecular recognition between highly charged molecules such as nucleic acids, inorganic ions, charged amino acids, etc. For such cases, it has been demonstrated that molecular simulations with fixed partial charges often fail to achieve experimental accuracy. Although incorporation of more advanced electrostatic models (such as multipoles, mutual polarization, etc.) can significantly improve simulation accuracy, it increases computational expense by a factor of 5-20×. Indirect free energy (IFE) methods can mitigate this cost by modeling intermediate states at fixed-charge resolution. For example, an efficient "reference" model such as a pairwise Amber, CHARMM, or OPLS-AA force field can be used to derive an initial estimate, followed by thermodynamic corrections to a more advanced "target" potential such as the polarizable AMOEBA model. Unfortunately, all currently described IFE methods encounter difficulties reweighting more than ∼50 atoms between resolutions due to extensive scaling of both the magnitude of the thermodynamic corrections and their statistical uncertainty. We present an approach called "simultaneous bookending" (SB) that is fundamentally different from existing IFE methods based on a tunable sampling approximation, which permits scaling to thousands of atoms. SB is demonstrated on the relative binding affinity of Mg2+/Ca2+ to a set of metalloproteins with up to 2972 atoms, finding no statistically significant difference between direct AMOEBA results and those from correcting Amber to AMOEBA. The ability to change the resolution of thousands of atoms during reweighting suggests the approach may be applicable in the future to protein-protein binding affinities or nucleic acid thermodynamics.