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.     

纳米接触行为的三维多尺度数值模拟

应用分子动力学与有限元耦合的桥域多尺度算法,模拟三维刚性球形压头与光滑基体表面的纳米尺度接触行为,并与全原子分子模拟结果比较.考察在一定载荷下的系统弛豫行为、两种模型桥接区位移和应力的连续性、法向力和接触面积随压头位移变化等,结果表明:一定外载荷下,桥域多尺度算法能较快达到平衡状态,且压头的振荡幅度更小,系统初始温度为0 K时该算法的相对误差最小.在准静态加载过程中,该算法能够将原子区的位移、应力等连续的过渡到连续介质区,具有较好的耦合效果;法向力-压头位移和接触半径-压头位移曲线几乎与分子模拟结果重合,表明算法具有较高的计算精度.     

粗粒化模型中的Gay-Berne相互作用势

合理描述非球形粗粒化粒子间的相互作用是提高粗粒化分子动力学模拟速度的关键.为此本文介绍了Gay-Berne势.将之应用于两种有机小分子体系,在合理选择构象集后,由遗传算法得到了粗粒化粒子的GB参数,并通过对粗粒化模型和全原子模型得到的范德瓦耳斯相互作用的对比检验了GB力场参数.最后,指出如何处理作用位点是粗粒化模型发展的一个关键问题.     

随机反应扩散格气模型上钙动力学的粗粒化模拟

发展了一种基于随机格气模型的粗粒化方法,该方法能有效模拟内质网表面钙动力学信息. 首先将相邻的微观节点合并成粗粒化节点,再根据局域平均场近似推导出粗粒化反应速率,然后执行粗粒化动力学蒙特卡洛模拟. 发现粗粒化动力学蒙特卡洛模拟结果和微观模拟结果非常吻合. 有趣的是,存在一个最佳的粗粒化比m,使得粗粒化模拟与微观模拟的相变点偏差最小. 固定m,发现临界点随体系尺度增加而单调增加,而且相变点的偏差与体系尺度存在一个标度关系.此外,该粗粒化方法大大地加快了蒙特卡洛模拟速率,并且与微观模拟直接相关. 该方法可以广泛用来研究体系尺度效应,而节省大量计算时间.     

硅功能化石墨烯负极材料的粗粒模型

硅功能化石墨烯(硅化烯)作为锂离子电池的负极材料, 一旦发生分层或粉化等损伤现象, 会严重地降低材料的电子输运能力和储锂容量, 减少电池的使用寿命, 因此要求负极材料具有较强的力学可靠性. 考虑到传统分子动力学方法的模拟尺度很难达到硅化烯负极材料的真实尺度, 首先采用Tersoff 势函数和Lennard-Jones 势函数建立了多种硅化烯的全原子数值模型, 计算材料的各种弹性模量和吸附能; 然后采用珠子-弹簧结构, 根据力学平衡条件和能量守恒定律, 结合全原子模型的计算结果, 建立了硅化烯粗粒模型及其系统的能量方程; 最后, 通过对比石墨烯粗粒模型与其全原子模型的拉伸性能, 验证了硅化烯粗粒模型的有效性.     

DNA超分子水凝胶的粗粒化建模与模拟

DNA超分子水凝胶在生物、医学领域具有广阔的应用前景,基于计算模拟技术研究其分子结构与其宏观性能关系具有重要意义.归因于其复杂的结构和较大的相关时间空间尺度,目前针对DNA水凝胶的分子建模与模拟研究比较缺乏.本文建立了一种DNA水凝胶的粗粒化模型,采用分子动力学模拟的方法针对一种模块化纯DNA水凝胶进行了研究.模拟结果表明其凝胶态介观结构为多孔海绵状,其交联度与水凝胶的浓度成正相关,并得出了其转变温度的范围等.模拟结果与相关实验定性或半定量符合,表明该模型可能用于该模块化DNA水凝胶等类似系统的结构功能关系研究.     

磷脂微滴囊泡化的介观模拟

发展了一种非显示溶剂的粗粒化三粒子磷脂模型,该模型明确反映磷脂分子的双尾结构．模型分别采用变形的MIE作用势和Harmonic作用势描述分子间非成键和分子内成键相互作用,粗粒化力场参数通过拟合DPPC双分子层的结构和力学性质获得．该粗粒化模型成功重现了磷脂分子从随机初始态到双分子层和从盘状结构到囊泡的形成过程．应用该模型系统研究了球形和柱形磷脂微滴囊泡化的过程,结果表明此模型能有效地模拟介观尺度下复杂磷脂囊泡的形成及演化．     

生物分子模拟中的静电计算

在分子尺度上对生物系统的研究主要通过模拟生物分子间的相互作用来完成.在众多的分子间相互作用中,静电相互作用及其重要,对它的定量计算是了解生物分子的溶剂化效应、生物分子折叠和酶催化作用等的一个中心课题.本综述着重介绍用于计算溶剂化效应的泊松-玻尔兹曼模型,包括模型的理论基础、发展历程、数值求解方法、及在分子生物学研究中的应用等.同时,也介绍泊松-玻尔兹曼模型的一种近似模型,广义玻恩模型.     

气固鼓泡床的全尺度直接数值模拟

颗粒全尺度直接数值模拟由于不引入任何相间作用力封闭模型、同时又考虑了颗粒相与流体相之间的四向耦合,因而具有极高的计算精度和准度。本文采用内嵌边界多重直接力算法结合软球碰撞模型对实验室尺度的气固鼓泡流化床装置进行了全尺度直接数值模拟。计算结果真实还原了鼓泡流态化的流动情形,且能有效地捕捉装置内部流场的详细运动情况和涡结构。对流化床内颗粒在传统的离散元模型框架下进行曳力统计发现,平均曳力比全尺度模拟结果小约20%~30%。其具体数值因选取比较的曳力模型和统计网格而异。改进传统曳力模型需考虑颗粒群的非均匀性以及颗粒拟温度。     

软物质实验方法前沿:单分子操控技术

传统的分子生物学实验方法基本都是系综的方法,测量的信号来自大量的生物分子的平均响应,这不利于得到生物分子的构象转变与功能的动力学细节.另外,很多生物大分子如细胞骨架蛋白、分子马达等在行使功能的时候都会受到或者产生力的作用,传统的实验方法也难于研究生物分子的力学响应.最近20年左右发展起来的单分子操控技术可以实现对单个分子的操控,同时测量单个分子在拉力作用下的力学响应.最为常用的单分子操控技术主要包括光镊、磁镊和原子力显微镜,不同的技术有不同的特点和适用范围.本文对几种常用的单分子操纵技术的特点,包括物理原理、可以施加的力的范围与精度、可以测量的分子长度范围与精度等做一个系统的介绍.另外,单分子操控技术在生物大分子如核糖核酸(DNA),脱氧核糖核酸(RNA)和蛋白质的构象转变,相互作用,以及分子马达的功能机理等方面已经取得的丰富成果也给出概括性的介绍.本文对生物学家系统的了解单分子操控技术和如何应用该技术解决自己的生物问题提供一个有益的参考.     

Structure-based simulations complemented by conventional all-atom simulations to provide new insights into the folding dynamics of human telomeric G-quadruplex

The hybrid atomistic structure-based model has been validated to be effective in investigation of G-quadruplex folding. In this study, we performed large-scale conventional all-atom simulations to complement the folding mechanism of human telomeric sequence Htel24 revealed by a multi-basin hybrid atomistic structure-based model. Firstly, the real time-scale of folding rate, which cannot be obtained from the structure-based simulations, was estimated directly by constructing a Markov state model. The results show that Htel24 may fold as fast as on the order of milliseconds when only considering the competition between the hybrid-1 and hybrid-2 G-quadruplex conformations. Secondly, in comparison with the results of structure-based simulations, more metastable states were identified to participate in the formation of hybrid-1 and hybrid-2 conformations. These findings suggest that coupling the hybrid atomistic structure-based model and the conventional all-atom model can provide more insights into the folding dynamics of DNA G-quadruplex. As a result, the multiscale computational framework adopted in this study may be useful to study complex processes of biomolecules involving large conformational changes.     

A multiscale coupling method for the modeling of dynamics of solids with application to brittle cracks

We present a multiscale model for numerical simulations of dynamics of crystalline solids. The method combines the continuum nonlinear elasto-dynamics model, which models the stress waves and physical loading conditions, and molecular dynamics model, which provides the nonlinear constitutive relation and resolves the atomic structures near local defects. The coupling of the two models is achieved based on a general framework for multiscale modeling – the heterogeneous multiscale method (HMM). We derive an explicit coupling condition at the atomistic/continuum interface. Application to the dynamics of brittle cracks under various loading conditions is presented as test examples.     

Non-Equilibrium Behaviour in Unidirectionally Coupled Map Lattices

A spatially one dimensional coupled map lattice with a local and unidirectional coupling is introduced. This model is studied analytically by a perturbation theory that is valid for small coupling strength. In parameter space three phases with different ergodic behaviour are observed. Via coarse graining the deterministic model is mapped to a stochastic spin model that can be described by a master equation. Because of the anisotropic coupling non-equilibrium behaviour is found on the coarse grained level. However, the stationary statistical properties of the spin dynamics can still be described with a nearest neighbour Ising model whereby the ordering is predominantly antiferromagnetic.     

Embedding atom-jellium model for metal surface

To describe metal surfaces efficiently and accurately, an embedding atom-jellium model is proposed. Within density functional theory, we consider a multiscale scheme that combines jellium and atomistic approaches. We use the former to model layers deep inside a metal surface to reduce the computational cost and the later to maintain the accuracy required for chemical bonding. Work functions of Al(111) and Cu(111) surfaces are studied using this model with comparisons to all-atom and pure jellium models. The much closer results of the embedding atom-jellium model to the all-atom results than to the pure jellium results show a good prospect for our approach in large-scale density functional calculations.     

Local–global multiscale model reduction for flows in high-contrast heterogeneous media

In this paper, we study model reduction for multiscale problems in heterogeneous high-contrast media. Our objective is to combine local model reduction techniques that are based on recently introduced spectral multiscale finite element methods (see [19]) with global model reduction methods such as balanced truncation approaches implemented on a coarse grid. Local multiscale methods considered in this paper use special eigenvalue problems in a local domain to systematically identify important features of the solution. In particular, our local approaches are capable of homogenizing localized features and representing them with one basis function per coarse node that are used in constructing a weight function for the local eigenvalue problem. Global model reduction based on balanced truncation methods is used to identify important global coarse-scale modes. This provides a substantial CPU savings as Lyapunov equations are solved for the coarse system. Typical local multiscale methods are designed to find an approximation of the solution for any given coarse-level inputs. In many practical applications, a goal is to find a reduced basis when the input space belongs to a smaller dimensional subspace of coarse-level inputs. The proposed approaches provide efficient model reduction tools in this direction. Our numerical results show that, only with a careful choice of the number of degrees of freedom for local multiscale spaces and global modes, one can achieve a balanced and optimal result.     

Variational multiscale turbulence modelling in a high order spectral element method

In the variational multiscale (VMS) approach to large eddy simulation (LES), the governing equations are projected onto an a priori scale partitioning of the solution space. This gives an alternative framework for designing and analyzing turbulence models. We describe the implementation of the VMS LES methodology in a high order spectral element method with a nodal basis, and discuss the properties of the proposed scale partitioning. The spectral element code is first validated by doing a direct numerical simulation of fully developed plane channel flow. The performance of the turbulence model is then assessed by several coarse grid simulations of channel flow at different Reynolds numbers.     

Adaptive resolution simulation of an atomistic DNA molecule in MARTINI salt solution

We present a dual-resolution model of a deoxyribonucleic acid (DNA) molecule in a bathing solution, where we concurrently couple atomistic bundled water and ions with the coarse-grained MARTINI model of the solvent. We use our fine-grained salt solution model as a solvent in the inner shell surrounding the DNA molecule, whereas the solvent in the outer shell is modeled by the coarse-grained model. The solvent entities can exchange between the two domains and adapt their resolution accordingly. We critically asses the performance of our multiscale model in adaptive resolution simulations of an infinitely long DNA molecule, focusing on the structural characteristics of the solvent around DNA. Our analysis shows that the adaptive resolution scheme does not produce any noticeable artifacts in comparison to a reference system simulated in full detail. The effect of using a bundled-SPC model, required for multiscaling, compared to the standard free SPC model is also evaluated. Our multiscale approach opens the way for large scale applications of DNA and other biomolecules which require a large solvent reservoir to avoid boundary effects.     

Soot primary particle formation from multiscale coarse-grained molecular dynamics simulation

A new multiscale coarse-graining procedure is used to study carbonaceous nanoparticle agglomeration in combustion environments. The computational methodology is applied to an ensemble of 10,000 nanoparticles (or effectively 2 million total carbon atoms) to simulate, for the first time, the agglomeration of carbonaceous nanoparticles using coarse-grained atomistic-scale information. In particular, with the coarse-graining approach we are able to assess the influence of nanoparticle morphology and temperature on the agglomeration process. The coarse-graining of the interparticle force field is accomplished applying a force-matching procedure to data obtained from trajectories and forces from all-atom MD simulations. The coarse-grained MD results show rich and varied clustering behaviors for different particle morphology and, in some cases, the formation of primary particles with a diameter around 15 nm are observed for the first time by molecular simulation techniques.