Affentranger Roman, Tavernelli Ivano, Di Iorio Ernesto E
Institut für Biochemie, Eidgenössische Technische Hochschule ETH-Zurich, Schafmattstrasse 18, 8093 Zurich, Switzerland, and Institut de Chimie Moléculaire et Biologique, BCH-LCBC, Ecole Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland.
J Chem Theory Comput. 2006 Mar;2(2):217-28. doi: 10.1021/ct050250b.
Limited searching in the conformational space is one of the major obstacles for investigating protein dynamics by numerical approaches. For this reason, classical all-atom molecular dynamics (MD) simulations of proteins tend to be confined to local energy minima, particularly when the bulk solvent is treated explicitly. To overcome this problem, we have developed a novel replica exchange protocol that uses modified force-field parameters to treat interparticle nonbonded potentials within the protein and between protein and solvent atoms, leaving unperturbed those relative to solvent-solvent interactions. We have tested the new protocol on the 18-residue-long tip of the P domain of calreticulin in an explicit solvent. With only eight replicas, we have been able to considerably enhance the conformational space sampled during a 100 ns simulation, compared to as many parallel classical molecular dynamics simulations of the same length or to a single one lasting 450 ns. A direct comparison between the various simulations has been possible thanks to the implementation of the weighted histogram analysis method, by which conformations simulated with modified force-field parameters can be assigned different weights. Interatom, inter-residue distances in the structural ensembles obtained with our novel replica exchange approach and by classical MD simulations compare equally well with those derived from NMR data. Rare events, such as unfolding and refolding, occur with reasonable statistical frequency. Visiting of conformations characterized by very small Boltzmann weights is also possible. Despite their low probability, such regions of the conformational space may play an important role in the search for local potential-energy minima and in dynamically controlled functions.
在构象空间中进行有限搜索是通过数值方法研究蛋白质动力学的主要障碍之一。因此，蛋白质的经典全原子分子动力学（MD）模拟往往局限于局部能量最小值，特别是当明确处理大量溶剂时。为了克服这个问题，我们开发了一种新颖的副本交换协议，该协议使用修改后的力场参数来处理蛋白质内部以及蛋白质与溶剂原子之间的粒子间非键合势，而相对于溶剂 - 溶剂相互作用的那些势则保持不变。我们在明确的溶剂中对钙网蛋白P结构域18个残基长的末端测试了新协议。与相同长度的多个并行经典分子动力学模拟或持续450 ns的单个模拟相比，仅使用八个副本，我们就能在100 ns模拟期间显著增强采样的构象空间。由于实施了加权直方图分析方法，各种模拟之间的直接比较成为可能，通过该方法，可以为用修改后的力场参数模拟的构象分配不同的权重。我们新颖的副本交换方法和经典MD模拟获得的结构集合中的原子间、残基间距离与从NMR数据得出的距离同样匹配良好。罕见事件，如展开和重新折叠，以合理的统计频率发生。访问具有非常小的玻尔兹曼权重特征的构象也是可能的。尽管其概率很低，但构象空间的这些区域可能在寻找局部势能最小值和动态控制功能中发挥重要作用。
