蛋白质是生物体内主要的功能性高分子化合物。由于生物体始终处于动态非平衡状态，许多环境因素如温度、pH、局部离子浓度、大分子拥挤环境以及各种配体蛋白等都会对蛋白质的结构及其运动变化产生影响。精确描述蛋白质在不同环境因素作用下的构象变化是理解各种生理过程机理的基础。传统分子生物学实验手段由于较低的时空分辨率，难以直接精确研究蛋白质构象的动态变化。理论计算特别是分子动力学模拟（Molecular dynamics simulation, MD）成为在原子尺度上研究蛋白质分子微观结构和动态变化特性的重要工具。本论文在扩展与优化课题组先前开发的MD软件包基础上，选取三个外界环境因素引发的蛋白质构象变化问题进行研究。重点分析了控制机制在竞争中的协调对蛋白质动态结构的影响。主要结果和结论如下：第二章对课题组先前开发的MD软件包GPU_MD-1.0.5进行了扩展与优化，引入更加常用和精确的CHARMM27力场。通过把长程力PME算法中需要全局通讯的步骤放到CPU端计算，实现了一种GPU+CPU异构PME算法，进一步提高了程序的计算效率。新的模拟软件包命名为GPU_MD-2.0。通过对多种不同体系的模拟，对程序准确性和性能进行了验证。与运行在CPU上的GROMACS-4.5.5相比，对于不同规模的模拟体系GPU_MD-2.0在单GPU卡上相比单线程GROMACS可达到约6~15倍加速比，相比8线程GROMACS可达到约1~2倍加速比。与GPU加速版本的GROMACS-4.6.2相比，在相同的硬件计算条件下，GPU_MD-2.0计算速度稍快，达到了较高的计算效率，为进行蛋白质结构变化的模拟打下了基础。第三章以剪切流场引发的血小板糖蛋白β-switch区域由无规则卷曲到β-折叠构象转变过程为例，研究外界流场引发的蛋白质构象变化机理。分别通过直接分子动力学、流动分子动力学以及Metadynamics等多种模拟手段对此构象转变过程进行了模拟和深入分析。尤其是通过基于路径联合变量（Path collective variable）的Metadynamics方法计算得到β-switch区域构象转变的完整自由能曲面。研究表明，在无流场作用下，体系自由能趋于最小是唯一控制机制，在其主导下β-switch区域呈现无规则卷曲状态。当引入外界流场后，流场作用引发了另一控制机制，即水分子以最小阻力通过蛋白质分子，在其主导下β-switch区域呈现β-折叠状态。两种控制机制在竞争中的协调导致了不同特征构象的交替出现，共同主导了β-switch区域构象的动态变化。第四章以固有无序蛋白α-MoRE与配体蛋白XD结合与折叠的耦合过程为例，研究结合引发的固有无序蛋白折叠机理。首先，采用并行回火（Parallel tempering）方法对孤立α-MoRE体系进行模拟；其次，采用Metadynamics与并行回火相结合的增强取样方法对α-MoRE与XD结合与折叠的耦合过程在显式溶剂中进行了全原子模拟并得到其自由能曲面。研究表明，对于孤立α-MoRE，体系自由能趋于最小是唯一控制机制，在其主导下α-MoRE呈无规则卷曲和部分折叠共存的状态。当配体蛋白XD存在时，α-MoRE与XD的结合作用引发了另一控制机制，即结合能趋于最小，在其主导下α-MoRE呈现完整α-螺旋构象。两种控制机制在竞争中的协调共同主导了α-MoRE与XD结合与折叠的耦合过程。第五章对一个更加复杂的固有无序蛋白，肿瘤抑制蛋白p53碳端结构域片段（p53 CTD）进行模拟研究。首先，采用并行回火方法对孤立p53 CTD体系构象系综进行充分取样；其次，采用直接分子动力学分别对三个p53 CTD与不同配体蛋白相结合的复合物体系进行模拟研究，以考察配体蛋白存在时体系自由能与结合作用的相互关系。研究表明，对于孤立p53 CTD，体系自由能趋于最小是唯一控制机制，在其主导下p53 CTD呈无规则卷曲状态，但有一定形成二级结构的倾向。当配体蛋白存在时，来自于不同配体蛋白的结合作用引发了另一控制机制，即结合能趋于最小，在其主导下p53 CTD呈现相应的结合态构象。两种控制机制在竞争中的协调导致了不同特征构象的交替出现，而两者的相对强弱则决定了各种特征构象的取样频率。随着计算机模拟软硬件的持续发展以及各种先进增强取样方法的进一步完善，更多环境因素作用下蛋白质构象变化问题可以得到建模与研究，MD体系设置应更多考虑生物体内的复杂环境。在这些工作的基础上，蛋白质结构变化的控制机制与稳定性条件也将得到进一步研究与阐述。 

Proteins are the major functional macromolecules in organisms. Since the organism is always in dynamic non-equilibrium states, various external environments like temperature, pH, local ion concentration, macromolecular crowding and binding interaction from partners are all important factors which influence the protein structure and dynamics. Accurate description of the dynamics of proteins under different environments is the basis to understand the mechanisms of various physiological processes. Due to the limited spatio-temporal resolution, traditional experimental methods are difficult to study the dynamic changes of proteins directly. Computational simulation approach, especially molecular dynamics (MD) simulation, constitutes a valuable tool to study such problems on the atomic scale. In this work, previously developed MD simulation package is extended and optimized. On this basis, three environment induced protein conformational change problems are investigated. The compromise in competition between the underlying mechanisms and their effects on protein dynamic structures is analyzed. The main results and conclusions are as follows:In chapter 2, previously developed MD simulation package, GPU_MD-1.0.5, was extended and optimized. The more accurate and commonly used force-field, CHARMM27, was integrated into the package. In addition, a GPU+CPU heterogeneous PME algorithm was developed through putting the global communication steps on CPU in the PME algorithm. The new simulation package was named GPU_MD-2.0. Several systems were simulated to validate the accuracy and performance of the package. Compared with the GROMACS-4.5.5 running on CPU, GPU_MD-2.0 on a single GPU card could achieve about 6~15 times speed up for single thread GROMACS and about 1~2 times speed up for 8 threads GROMACS. Compared with the GPU accelerated GROMACS-4.6, the computing speed of GPU_MD-2.0 is slightly faster under the same hardware conditions, illustrating that GPU_MD-2.0 has achieved relatively high computing efficiency.In chapter 3, a flow induced protein conformational transition process, that is, the external flow induced loop to β-sheet conformational change in the β-switch region of glycoprotein Ibα was investigated. Direct MD, flow MD and metadynamics were employed to investigate the mechanisms of this flow induced conformational transition process. Specifically, the free energy landscape of the whole transition process was calculated by metadynamics with the path collective variable approach. The results revealed that without external flow, the free energy tending to the minimum is the solely dominant mechanism and the β-switch adopts random coil conformations. When the external flow exists, the dynamics of β-switch is dominated jointly by two mechanisms, the free energy tending to the minimum and water molecules tending to flow through the protein with minimum resistance. Each of these mechanisms has an extreme tendency that corresponds to a possible characteristic state. The compromise in competition between these two mechanisms leads to alternate occurrence of different characteristic states, resulting in the dynamics structures of β-switch.In chapter 4, the coupled folding and binding process of intrinsically disordered protein (IDP), α-MoRE, to its partner protein XD was investigated. The isolated α-MoRE system was first simulated by the parallel tempering method. Then the coupled folding and binding process of α-MoRE to XD was simulated by a combined metadynamics and parallel tempering in explicit solvent. Starting from an unbound and partially folded state of α-MoRE, multiple folding and binding events were observed during the simulation and the energy landscape was well estimated. The results revealed that for isolated α-MoRE, the free energy tending to the minimum is the solely dominant mechanism, and α-MoRE adopts random coil or partially folded conformations. When the partner protein XD is present, the binding effect constitutes another dominant mechanism, that is, the binding energy tending to the minimum and the α-MoRE tends to adopt fully helical conformation. The compromise in competition between these two mechanisms jointly determines the coupled folding and binding process of α-MoRE to XD.In chapter 5, a more complicated IDP system, the C-terminal domain of tumor suppressor p53 (p53 CTD) was investigated. The isolated p53 CTD was first simulated by the parallel tempering method. The three regulatory binding complexes of p53 CTD with different partner proteins were simulated by direct MD to investigate the interaction between free energy and the binding effect. The results demonstrated that, for isolated p53 CTD system, the free energy tending to the minimum is the solely dominant mechanism, and p53 CTD mainly adopts random coil conformations with a limited extent to form helical structures. However, when the binding partners present, the binding interactions constitute another dominant mechanism, that is, the binding energy tending to the minimum and the p53 CTD tends to adopt the corresponding bound structures. The compromise in competition between these two mechanisms leads to alternate occurrence of different characteristic states, and the relative strength of the two mechanisms determines the sampling frequency of each state.With the continuous development of computer hardware and software, as well as further improvement of various advanced enhanced sampling methods, more protein conformational change problems induced by external environments could be modeled and investigated. MD simulations should also be more in line with the complex environment in living organisms. Based on these researches, the underlying mechanisms of protein structural changes and the stability conditions will be further investigated and elaborated.