增强采样分子动力学模拟在生物分子研究中的进展（英文）

分子动力学模拟能够描述蛋白质分子在行使生物学功能过程中涉及的构象变化,已发展成为中物学研究中重要的计算工具.由于生物分子的构象分布存在崎岖的自由能面,在较为复杂的生物体系的模拟中,传统的分子动力学模拟的构象采样能力受到极大限制,模拟的时间尺度与真实的生物学过程之间仍存在差距.增强采样是解决这一问题的有效手段.本文综述了两类增强采样方法即约束型和无约束型增强采样算法的理论基础、最新进展及其在生物分子中的典型应用,同时也简要总结了组合型增强采样算法近些年的发展.     

基于TorchMD的粗粒化分子动力模拟研究

粗粒化模型通过简化原子性质以及原子间的相互作用实现生物大分子长时间尺度的分子动力学模拟. 深度学习通过模拟人类的认知过程实现海量数据的准确分类和回归过程. 本论文将这两种技术进行融合,利用基于深度学习的粗粒化分子动力学模拟技术研究分子在不同状态之间的变化过程,并提出基于TorchMD的分子动力学模拟的分析框架. 在本工作中,MFDP聚类算法被用于在三维的CV变量空间中进行聚类,并确定分子的若干主要状态,在完成聚类的同时,给出各类中的代表分子构象,并给出类之间的分子构象. 这为后续利用String算法分析分子在不同状态间的转换路径打下基础. 通过String算法,迭代搜索得到分子在不同状态之间的变化路径以及对应的势能变化曲线. 通过与已有文献的结果进行对比,验证了基于TorchMD的粗粒化分子动力学模拟的理论框架可以在相对较短的时间尺度里研究分子的变化过程.     

基于接触数引导的迭代多组独立分子动力学模拟实现配体-蛋白质结合与解离过程

配体的结合与解离过程在蛋白质实现其生物学功能方面非常关键,因此对这些高度动态过程的研究变得非常重要. 尽管已有实验方法可以确定蛋白质-配体复合物的三维结构,但一般仅可获得静态图片. 随着计算机算力的快速提高以及算法的优化,分子动力学模拟在探索配体的结合与解离过程方面具有诸多优势. 然而,当系统变得足够大时,分子动力学模拟的时间和空间尺度成为了巨大的挑战. 本工作提出了一种研究配体-蛋白质结合与解离的增强采样工具,它基于配体和蛋白质之间形成的接触数来引导迭代多组独立分子动力学模拟. 在腺苷酸激酶的模拟结果中,观测到配体的结合和解离过程,而使用传统分子动力学模拟在同一时间尺度下则无法实现这一过程.     

分析复杂化学及生物体系分子动力学模拟轨迹的聚类方法（英文）

分子动力学(MD)模拟可以很好地用于揭示蛋白质等生物大分子体系在原子尺度的结构及功能的关系.分子动力学模拟通常产生海量的描述分子在模拟中运动的数据,包含很多模拟轨迹以及随时间演化的各个原子的坐标和速度等.为了从这些海量数据中获得体系的分子机制,需要发展并利用聚类算法来将这些海量数据进行归类,聚类算法通常将具有某些相似度的构象聚成一类,这些相似度可以分为两类,几何相似度以及动力学相似度.对应地,用于分析分子动力学模拟的聚类算法通常可以分为两大类:几何聚类及动力学聚类.本文列举了一系列常用的用于分子动力学模拟的聚类算法包括分裂算法,凝聚算法(单连锁,完全连锁,平均连锁,质心连锁以及Ward连锁),中心算法(K-Means,KMedoids,K-Centers及APM),密度算法(邻居算法,DBSCAN,密度-峰及Robust-DB算法),谱算法(PCCA, PCCA+)等.本文讨论了几何分类和动力学分类的不同点以及不同算法的性能.另外注意到并不存在某一个适用于所有MD数据的聚类算法.对于某个特定体系,选择一个合适的聚类算法取决于聚类的目的,MD构象系综的内在性质等.因此,本文的一个要点也在于介绍每个聚类算法的优缺点.期望通过本文,能够指导读者在MD模拟中选择一个合适的聚类算法.     

甘氨酸在纳米碳管中的吸附及性质的分子模拟

采用分子力学、分子动力学方法模拟研究了甘氨酸分子在单壁纳米碳管中的吸附和扩散行为 ,并对甘氨酸分子在纳米碳管中的构象和能量进行了优化 .模拟计算结果表明 ,甘氨酸在纳米碳管中的构象发生了伸缩和扭转 ,这种构象的改变将会导致氨基酸生物性能的改变 ;纳米碳管对氨基酸分子具有较强的吸附作用 ,其中纳米碳管和甘氨酸分子之间的π -π相互作用增加了纳米碳管对氨基酸的吸附能 .模拟过程中氨基酸分子和纳米碳管之间的运动会保持很强的协同效应 ,使模拟体系构型在能量上处于最稳定的状态     

二维NOESY核磁共振波谱结合全原子分子动力学模拟研究肌肽在水溶液中的构象变化

采用二维NOESY核磁共振波谱结合全原子分子动力学模拟研究了肌肽在水溶液中的构象变化和相互作用. 以分子内距离、回转半径、RMSD以及溶剂可接触表面积等性质进行表征. 分子动力学模拟显示肌肽分子在水溶液中表现出了较高的柔性,其构象在伸展、折叠之间互相转换,大部分情况下是以伸展的构象为主导的,而折叠构象较少. 二维NOESY核磁共振实验证实了模拟的结果,实验与理论得到很好的吻合.     

基于接触数引导的迭代多组独立分子动力学模拟实现配体-蛋白质结合与解离过程（英文）

配体的结合与解离过程在蛋白质实现其生物学功能方面非常关键,因此对这些高度动态过程的研究变得非常重要.尽管已有实验方法可以确定蛋白质-配体复合物的三维结构,但一般仅可获得静态图片.随着计算机算力的快速提高以及算法的优化,分子动力学模拟在探索配体的结合与解离过程方面具有诸多优势.然而,当系统变得足够大时,分子动力学模拟的时间和空间尺度成为了巨大的挑战.本工作提出了一种研究配体-蛋白质结合与解离的增强采样工具,它基于配体和蛋白质之间形成的接触数来引导迭代多组独立分子动力学模拟.在腺苷酸激酶的模拟结果中,观测到配体的结合和解离过程,而使用传统分子动力学模拟在同一时间尺度下则无法实现这一过程.     

粗粒化自由能模拟研究ABC输出蛋白质构象态转变

本文应用了粗粒化分子动力学方法模拟计算了ATP结合盒式输出蛋白沿其构象转变途径反应坐标的平均力势，这个反应坐标被定义为内门和外门质量中心距离之差. 计算得到的平均力势能很好地描述不同的向内构象态、向外构象态和阻塞构象状态，以及它们之间的转变. 粗粒化分子动力学自由能模拟显示，在向内构象态到向外构象态转变过程中，内门在外门打开之前先行关闭；反之，在向外构象态到向内构象态转变过程中，外门在内门大开之前先行关闭. 因此，在向内构象态和向外构象态两种转变过程中，都经过了阻塞构象状态. 模拟结果揭示了ATP结合盒式输出蛋白的传输单向性，这种特性在生命体系功能实现中具有十分重要的意义. 这些结果与先前晶体结构实验［Proc. Natl. Acad. Sci. USA 104，19005 (2007)］发现有根本的不同，这些实验结果显示了内外门同时打开的不合理结果. 本文通过计算模拟阐明了ABC输出蛋白构象态变化的分子机理.     

噬菌体T4溶菌酶溶液构象模拟中分子力场和水模型的选择

蛋白质在溶液中可能以不同构象的集合形式存在,不能用单一的静态结构来表示. 分子动力学模拟已成为对溶液中蛋白质构象进行采样的有用工具,但分子力场和水模型的选择是关键问题. 这项工作介绍了噬菌体T4溶菌酶的个例研究. 本文发现,使用经典的AMBER99SB力场和TIP4P水模型,分子动力学模拟不能很好地描述野生型噬菌体T4溶菌酶在微秒时间尺度上的铰链弯曲结构域运动. 其它新型力场和水模型的组合,如被称为RSFF2+的残基特异性力场和离散校正的水模型TIP4P-D,能够对噬菌体T4溶菌酶溶液构象进行合理的采样,与实验数据有良好的一致性. 这项工作为进一步研究噬菌体T4溶菌酶的溶液构象转变提供了分子力场和水模型的参考.     

分子构象聚类中的K-means密度峰值搜索算法

分子构象的聚类是搜索分子动力学模拟轨迹中代表构象的主要方法。 它是分析复杂构象改变或分子间相互作用机制的关键步骤. 作为一种基于密度的聚类算法,密度峰值搜索算法因其聚类的准确度而被应用于分子聚类过程中. 但随着模拟时长的增长,密度峰值搜索算法较低的计算效率限制了其应用的可能. 本文提出K-means密度峰值搜索算法的聚类算法,它是密度峰值搜索算法在计算效率方面的一个扩展版本,用于解决密度峰值搜索算法中巨大的资源消耗问题. 在K-means密度峰值搜索算法中,首先,通过高效的聚类算法(例如K-means)进行初始聚类,得到的聚类中心被定义为具有权重的典型点. 然后,对加权的典型点通过密度峰值搜索算法实现二次聚类,并细化点为核心点、边界点、加细光晕点. 在与密度峰值搜索算法具有相似的精度的同时,计算复杂度由O(n²)降至O(n). 通过二面角,二级结构,关联图描述的分子构象,将KFDP用于多个模拟轨迹的聚类过程中. 并通过与K-means聚类算法,DBSCAN聚类算法的比较结果,验证了K-means密度峰值搜索算法的优势.     

Ligand diffusion in proteins via enhanced sampling in molecular dynamics

Computational simulations in biophysics describe the dynamics and functions of biological macromolecules at the atomic level. Among motions particularly important for life are the transport processes in heterogeneous media. The process of ligand diffusion inside proteins is an example of a complex rare event that can be modeled using molecular dynamics simulations. The study of physical interactions between a ligand and its biological target is of paramount importance for the design of novel drugs and enzymes. Unfortunately, the process of ligand diffusion is difficult to study experimentally. The need for identifying the ligand egress pathways and understanding how ligands migrate through protein tunnels has spurred the development of several methodological approaches to this problem. The complex topology of protein channels and the transient nature of the ligand passage pose difficulties in the modeling of the ligand entry/escape pathways by canonical molecular dynamics simulations. In this review, we report a methodology involving a reconstruction of the ligand diffusion reaction coordinates and the free-energy profiles along these reaction coordinates using enhanced sampling of conformational space. We illustrate the above methods on several ligand–protein systems, including cytochromes and G-protein-coupled receptors. The methods are general and may be adopted to other transport processes in living matter.     

Machine learning for collective variable discovery and enhanced sampling in biomolecular simulation

Classical molecular dynamics simulates the time evolution of molecular systems through the phase space spanned by the positions and velocities of the constituent atoms. Molecular-level thermodynamic, kinetic, and structural data extracted from the resulting trajectories provide valuable information for the understanding, engineering, and design of biological and molecular materials. The cost of simulating many-body atomic systems makes simulations of large molecules prohibitively expensive, and the high-dimensionality of the resulting trajectories presents a challenge for analysis. Driven by advances in algorithms, hardware, and data availability, there has been a flare of interest in recent years in the applications of machine learning – especially deep learning – to molecular simulation. These techniques have demonstrated great power and flexibility in both extracting mechanistic understanding of the important nonlinear collective variables governing the dynamics of a molecular system, and in furnishing good low-dimensional system representations with which to perform enhanced sampling or develop long-timescale dynamical models. It is the purpose of this article to introduce the key machine learning approaches, describe how they are married with statistical mechanical theory into domain-specific tools, and detail applications of these approaches in understanding and accelerating biomolecular simulation.     

蛋白质空间结构的实验技术和理论方法

文章主要介绍几种蛋白质空间结构的实验测定方法，在现代生物学研究中，最常用的方法包括X射线晶体学、二维核磁共振（2D-NMR）和低温冷冻电镜，近几年发展起来的单分子技术在生物大分子动态结构的研究中应用越来越多，这些方法都有它们特定的时间和空间分辨率，所测定的结构及其动力学受环境热运动涨落的影响也非常不同，文章对这些问题作了较详细的分析，在蛋白质结构的理论方法方面，介绍了一个新的折叠理论及其与现有折叠模型的关系．讨论了模拟计算在研究蛋白质构象变化和动力学方面的应用，同时强调了分子动力学和蒙特卡罗方法．指出粗粒化模型是研究的热点之一，对生物学中经常遇到的多长度多时问尺度问题提供了一个可行的解决方案。     

Low-temperature behavior of water confined by biological macromolecules and its relation to protein dynamics

Confined water is an essential component of biological entities and processes and its properties differ from the ones of bulk water. Since protein and water dynamics are thought to be strongly coupled, and since macromolecular dynamics is crucial for biological function, the study of water confined by biological macromolecules is not only interesting on its own right but often provides useful information for understanding biological activity at the molecular level. Studies are reviewed that focus on the low-temperature behavior of water confined in protein crystals and in stacks of native biological membranes. Diffraction methods allowed the determination of characteristic changes that relate to the glass transition and crystallization of water. Protein crystallography and energy-resolved neutron scattering are employed to gain further insight into the coupling of solvent and protein dynamics.Received: 1 January 2003, Published online: 14 October 2003PACS: 87.15.He Dynamics and conformational changes - 87.50.Gi Ionizing radiations (ultraviolet, X-rays, gamma-rays, ions, electrons, positrons, neutrons, and mesons, etc.) - 64.70.Pf Glass transitions     

Using novel variable transformations to enhance conformational sampling in molecular dynamics

One of the computational "grand challenges" is to develop methodology capable of sampling conformational equilibria in systems with rough energy landscapes. Here, a significant step forward is made by combining molecular dynamics with a novel variable transformation designed to enhance sampling by reducing barriers without introducing bias and, thus, to preserve, perfectly, equilibrium properties.     

Particle dispersion on rapidly folding random heteropolymers

We investigate the dynamics of a particle moving randomly along a disordered heteropolymer subjected to rapid conformational changes which induce superdiffusive motion in chemical coordinates. We study the antagonistic interplay between the enhanced diffusion and the quenched disorder. The dispersion speed exhibits universal behavior independent of the folding statistics. On the other hand it is strongly affected by the structure of the disordered potential. The results may serve as a reference point for a number of translocation phenomena observed in biological cells, such as protein dynamics on DNA strands.     

Replica-exchange method using the generalized effective potential

We propose an effective scheme for fast conformational searches by combining the replica exchange method (REM) with the generalized effect potential concept. The present method introduces the "q" value from the effective potential as a coupling parameter. It is found that the new method not only requires a much smaller number of replicas than the conventional REM, but also makes it possible to perform effective conformational sampling of complex systems with correct distributions maintained. The advantage of the present method has been demonstrated with in vacuo alanine dipeptide using a molecular dynamics simulation.     

In-Silico Study on the Interaction of Saffron Ligands and Beta-Lactoglobulin by Molecular Dynamics and Molecular Docking Approach

Safranal, crocetin, and dimethylcrocetin are secondary metabolites found in saffron and have a wide range of biological activities. An investigation of their interaction with a transport protein, such as β-lactoglobulin (β-lg), at the atomic level could be a valuable factor in controlling their transport to biological sites. The interaction of these ligands and β-lg as a transport protein was investigated using molecular docking and molecular dynamics (MD) simulation methods. The molecular docking results showed that safranal and crocetin bind on the surface of β-lg. However, dimethylcrocetin binds in the internal cavity of β-lg. The β-lg affinity for binding saffron ligands decreases in the following order: crocetin > dimethylcrocetin > safranal. The analysis of MD simulation trajectories showed that the β-lg and β-lg–ligand complexes became stable at approximately 3000 ps and that there was little conformational change in the β-lg–safranal and β-lg–dimethylcrocetin complexes over a 10-ns timescale. In addition, the profiles of atomic fluctuations showed the rigidity of the ligand binding site during the simulation time.     

Enhanced sampling based on slow variables of trajectory mapping

Most current enhanced sampling(ES) algorithms attempt to bias a potential energy surface based on preset slow collective variables to improve simulation efficiency. However, due to difficulty in obtaining slow variables in complex molecular systems,approximate slow variables are usually applied in ES, which often fail to achieve the expected high efficiency and sufficient accuracy when reconstructing equilibrium properties. In this paper, we demonstrate that the trajectory mapping(TM) technique has the potential to provide the required slow variables for ES. We illustrate the application of a typical ES algorithm(metadynamics)based on the slow variables constructed from the TM in a hairpin peptide system. In this system, both the equilibrium properties and slow dynamics are accurately obtained within approximately two to three orders of magnitude shorter simulation time than in regular molecular dynamics simulation.