Analysis of Long Molecular Dynamics Simulations Using Interactive Focus+Context Visualization

Analyzing molecular dynamics (MD) simulations is a key aspect to understand protein dynamics and function. With increasing computational power, it is now possible to generate very long and complex simulations, which are cumbersome to explore using traditional 3D animations of protein movements. Guided by requirements derived from multiple focus groups with protein engineering experts, we designed and developed a novel interactive visual analysis approach for long and crowded MD simulations. In this approach, we link a dynamic 3D focus+context visualization with a 2D chart of time series data to guide the detection and navigation towards important spatio‐temporal events. The 3D visualization renders elements of interest in more detail and increases the temporal resolution dependent on the time series data or the spatial region of interest. In case studies with different MD simulation data sets and research questions, we found that the proposed visual analysis approach facilitates exploratory analysis to generate, confirm, or reject hypotheses about causalities. Finally, we derived design guidelines for interactive visual analysis of complex MD simulation data.     

Denoising methods for thermomechanical decomposition for quasi-equilibrium molecular dynamics simulations

In this work, we applied robust denoising methods well established in the signal processing field for the thermomechanical decomposition of velocity data obtained from molecular dynamics (MD) simulations. In the decomposition, the atomic velocity was assumed to be the sum of the mechanical velocity and the thermal velocity, which can be linked to the stress and temperature field at the continuum scale, respectively. For the quasi-equilibrium process, with the thermal velocity treated as the Gaussian distributed stationary noise with zero mean and a positive variance that is linearly proportional to temperature and mechanical velocity as the clean signal, the velocity decomposition can be recasted into a denoising problem, for which powerful denoising methods have been developed to estimate a clean signal from noisy data. We investigated the widely-used linear parametric real-domain, linear nonparametric Fourier-based, and nonlinear nonparametric wavelet-based denoising methods, first on their theoretical properties and then made comparsion among them for denosing some synthetic noisy 1-dimensional (1D) data generated from MD simulations. The nonlinear wavelet-based thresholding estimator possessed better optimality properties than the other estimators, and also outperformed the other estimators in the synthetic data test. A further test comparing the various denoising methods for an adiabatic shear crack nucleation and propagation process simulated using MD simulations showed better performance by the wavelet-based denoisng method. Results from this work reveal good potential of applying wavelet-based denoising method to the study of thermomechanical processes simulated using MD simulations.     

Reaction analysis and visualization of ReaxFF molecular dynamics simulations

ReaxFF MD (Reactive Force Field Molecular Dynamics) is a promising method for investigating complex chemical reactions in relatively larger scale molecular systems. The existing analysis tools for ReaxFF MD lack the capability of capturing chemical reactions directly by analyzing the simulation trajectory, which is critical in exploring reaction mechanisms. This paper presents the algorithms, implementation strategies, features, and applications of VARxMD, a tool for Visualization and Analysis of Reactive Molecular Dynamics. VARxMD is dedicated to detailed chemical reaction analysis and visualization from the trajectories obtained in ReaxFF MD simulations. The interrelationships among the atoms, bonds, fragments, species and reactions are analyzed directly from the three-dimensional (3D) coordinates and bond orders of the atoms in a trajectory, which are accomplished by determination of atomic connectivity for recognizing connected molecular fragments, perception of bond types in the connected fragments for molecules or radicals, indexing of all these molecules or radicals (chemical species) based on their 3D coordinates and recognition of bond breaking or forming in the chemical species for reactions. Consequently, detailed chemical reactions taking place between two sampled frames can be generated automatically. VARxMD is the first tool specialized for reaction analysis and visualization in ReaxFF MD simulations. Applications of VARxMD in ReaxFF MD simulations of coal and HDPE (high-density polyethylene) pyrolysis show that VARxMD provides the capabilities in exploring the reaction mechanism in large systems with complex chemical reactions involved that are difficult to access manually.     

Effects of mutations on active site conformation and dynamics of RNA-dependent RNA polymerase from Coxsackievirus B3

Recent crystal structures of RNA-dependent RNA polymerase (3D^pol) from Coxsackievirus B3 (CVB3) revealed that a tyrosine mutation at Phe364 (F364Y) resulted in structures with open active site whereas a hydrophobic mutation at Phe364 (F364A) led to conformations with closed active site. Besides, the crystal structures showed that the F364W mutation had no preference between the open and closed active sites, similar to wild-type. In this paper, we present a molecular dynamics (MD) study on CVB3 3D^pol in order to address some important questions raised by experiments. First, MD simulations of F364Y and F364A were carried out to explore how these mutations at Phe364 influence active site dynamics and conformations. Second, MD simulations of wild-type and mutants were performed to discover the connection between active site dynamics and polymerase function. MD simulations reveal that the effect of mutations on active site dynamics is associated with the interaction between the structural motifs A and D in CVB3 3D^pol. Interestingly, we discover that the active site state is influenced by the formation of a hydrogen bond between backbone atoms of Ala231 (in motif A) and Ala358 (in motif D), which has never been revealed before.     

Three-dimensional finite element method for the filling simulation of injection molding

With the development of molding techniques, molded parts have more complex and larger geometry with nonuniform thickness. In this case, the velocity and the variation of parameters in the gapwise direction are considerable and cannot be neglected. A three-dimensional (3D) simulation model can predict the filling process more accurately than a 2.5D model based on the Hele–Shaw approximation. This paper gives a mathematical model and numeric method based on 3D model to perform more accurate simulations of a fully flow. The model employs an equal-order velocity–pressure interpolation method. The relation between velocity and pressure is obtained from the discretized momentum equations in order to derive the pressure equation. A 3D control volume scheme is used to track the flow front. During calculating the temperature field, the influence of convection items in three directions is considered. The software based on this 3D model can calculate the pressure field, velocity field and temperature field in filling process. The validity of the model has been tested through the analysis of the flow in cavities.     

Prediction of three-dimensional structures and structural flexibilities of wild-type and mutant cytochrome P450 1A2 using molecular dynamics simulations

In this study, we investigated the effect of genetic polymorphism on the three-dimensional (3D) conformation of cytochrome P450 1A2 (CYP1A2) using molecular dynamics (MD) simulations. CYP1A2, a major drug-metabolizing enzyme among cytochrome P450 enzymes (CYPs), is known to have many variant alleles. The genetic polymorphism of CYP1A2 may cause individual differences in the pharmacokinetics of medicines. By performing 100 ns or longer MD simulations, we investigated the influence of amino acid mutation on the 3D structures and the dynamic properties of proteins. The results show that the static structures were changed by the mutations of amino acid residues, not only near the mutated residues but also in distant portions of the proteins. Moreover, the mutation of only one amino acid was shown to change the structural flexibility of proteins, which may influence the substrate recognition and enzymatic activity. Our results clearly suggest that it is necessary to investigate the dynamic property as well as the static 3D structure for understanding the change of the enzymatic activity of mutant CYP1A2.     

Comparative numerical study of FEM methods solving gas damping in perforated MEMS devices

Three different numerical strategies are presented for the estimation of the damping force acting on perforated movable MEMS dampers. Results from the 2D Perforated Profile Reynolds (PPR) method and the simplified 2D ANSYS method are compared with accurate full 3D flow simulations. Altogether, 32 different topologies are compared varying, e.g., the dimensions of the square damper and the square holes, and the number of holes. The case of uniform perforation and perpendicular motion is studied. Oscillation in the low frequency regime is assumed, that is, the compressibility and inertia of the gas are ignored in the study. While the PPR method is in good agreement with the 3D simulations, the forces given by the ANSYS method were considerably smaller. The reasons for this are studied, and a compact expression to explain the small forces is derived.     

Three dimensional numerical simulations of long-span bridge aerodynamics, using block-iterative coupling and DES

The design of long-span bridges often depends on wind tunnel testing of sectional or full aeroelastic models. Some progress has been made to find a computational alternative to replace these physical tests. In this paper, an innovative computational fluid dynamics (CFD) method is presented, where the fluid-structure interaction (FSI) is solved through a self-developed code combined with an ANSYS-CFX solver. Then an improved CFD method based on block-iterative coupling is also proposed. This method can be readily used for two dimensional (2D) and three dimensional (3D) structure modelling. Detached-Eddy simulation for 3D viscous turbulent incompressible flow is applied to the 3D numerical analysis of bridge deck sections. Firstly, 2D numerical simulations of a thin airfoil demonstrate the accuracy of the present CFD method. Secondly, numerical simulations of a U-shape beam with both 2D and 3D modelling are conducted. The comparisons of aerodynamic force coefficients thus obtained with wind tunnel test results well meet the prediction that 3D CFD simulations are more accurate than 2D CFD simulations. Thirdly, 2D and 3D CFD simulations are performed for two generic bridge deck sections to produce their aerodynamic force coefficients and flutter derivatives. The computed values agree well with the available computational and wind tunnel test results. Once again, this demonstrates the accuracy of the proposed 3D CFD simulations. Finally, the 3D based wake flow vision is captured, which shows another advantage of 3D CFD simulations. All the simulation results demonstrate that the proposed 3D CFD method has good accuracy and significant benefits for aerodynamic analysis and computational FSI studies of long-span bridges and other slender structures.     

融合时序特征约束与联合优化的 点云3维人体姿态序列估计

目的 3维人体姿态估计传统方法通常采用单帧点云作为输入,可能会忽略人体运动平滑度的固有先验知识,导致产生抖动伪影。目前,获取2维人体姿态标注的真实图像数据集相对容易,而采集大规模的具有高质量3维人体姿态标注的真实图像数据集进行完全监督训练有一定难度。对此,本文提出了一种新的点云序列3维人体姿态估计方法。方法 首先从深度图像序列估计姿态相关点云,然后利用时序信息构建神经网络,对姿态相关点云序列的时空特征进行编码。选用弱监督深度学习,以利用大量的更容易获得的带2维人体姿态标注的数据集。最后采用多任务网络对人体姿态估计和人体运动预测进行联合训练,提高优化效果。结果 在两个数据集上对本文算法进行评估。在ITOP(invariant-top view dataset)数据集上,本文方法的平均精度均值(mean average precision,mAP)比对比方法分别高0.99%、13.18%和17.96%。在NTU-RGBD数据集上,本文方法的mAP值比最先进的WSM(weakly supervised adversarial learning methods)方法高7.03%。同时,在ITOP数据集上对模型进行消融实验,验证了算法各个不同组成部分的有效性。与单任务模型训练相比,多任务网络联合进行人体姿态估计和运动预测的mAP可以提高2%以上。结论 本文提出的点云序列3维人体姿态估计方法能充分利用人体运动连续性的先验知识,获得更平滑的人体姿态估计结果,在ITOP和NTU-RGBD数据集上都能获得很好的效果。采用多任务网络联合优化策略,人体姿态估计和运动预测两个任务联合优化求解,有互相促进的作用。     

Two- and three-dimensional lattice Boltzmann simulations of particle migration in microchannels

The use of two-dimensional (2D) numerical simulations with a reduced particle-based Reynolds number (Re) for studying particle migration in a microchannel with equally spaced multiple constrictions was investigated. 2D and 3D colloidal lattice Boltzmann (LB) models were used to simulate particle-fluid hydrodynamics. Experiments were conducted with inert microparticles in a creeping flow in a microflow channel with symmetric wall obstacles. Lowering Re in 2D simulations by a factor of R (the dimensionless particle radius in LB simulations) resulted in a close match between numerically computed and experimentally obtained particle velocities, indicating that Re-based dimensional scaling was needed to capture the 3D particle flow dynamics in 2D simulations of experimental data. We captured particle displacement motion in a microchannel with symmetric inline obstacles in 2D simulations, where symmetry in the flow field was broken by local disturbances in the flow field due to particle motion, indicating that asymmetry in channel geometry is not the sole cause for particle displacement motion. Particle acceleration/deceleration around each constriction followed the same pattern, but each constriction acted like a particle accelerator in 2D and 3D simulations, in which particles exhibited progressively higher velocities in each subsequent constriction. Particles migrated across multiple streamlines in converging and diverging flow zones in a creeping flow, which calls into question the use of steady streamlines for calculating transient particle flow. Monotonicity in particle acceleration toward the constriction and deceleration beyond the constriction was broken by interparticle hydrodynamic interactions leading to more pronounced particle migration across multiple streamlines.     

Predicting the structures of complexes between phosphoinositide 3-kinase (PI3K) and romidepsin-related compounds for the drug design of PI3K/histone deacetylase dual inhibitors using computational docking and the ligand-based drug design approach

Predictions of the three-dimensional (3D) structures of the complexes between phosphoinositide 3-kinase (PI3K) and two inhibitors were conducted using computational docking and the ligand-based drug design approach. The obtained structures were refined by structural optimizations and molecular dynamics (MD) simulations. The ligands were located deep inside the ligand binding pocket of the p110α subunit of PI3K, and the hydrogen bond formations and hydrophobic effects of the surrounding amino acids were predicted. Although rough structures were obtained for the PI3K–inhibitor complexes before the MD simulations, the refinement of the structures by these simulations clarified the hydrogen bonding patterns of the complexes.     

基于改进 L-K 光流的 WebAR 信息可视分析方法

摘 要：信息可视化技术结合移动增强现实(MAR)技术在目标跟踪领域仍然存在设备计算 负载过大的问题。若仍坚持采用同跟踪平面图像特征点的方案来跟踪立体对象各角度的特征点, 则目标跟踪过程所需要获取的多角度特征点数据无疑会加重跟踪过程的计算压力,进而导致移 动设备负载过大,最终影响模型渲染,所渲染的模型常出现剧烈抖动、卡顿或运动滞后于目标 物的现象。针对上述问题,提出了一种基于改进的 L-K (Lucas-Kanade)光流跟踪算法的 WebAR (基于 Web 端的 MAR 技术)解决方案,将特征点的跟踪问题转化为光流估计问题以及一种优化 的三维信息可视化交互策略。实验结果表明,该方法能够提高 MAR 在跟踪目标时的计算效率 和稳定性,丰富信息可视化的呈现效果和交互方式。     

双视三维重建的高精度运动参数估计方法

从两幅透视图像恢复被摄目标的三维结构是计算机视觉最基本的任务之一,其中,运动估计算法的性能决定了最终的三维重建精度。首先讨论了双视成像的基本数学模型,并介绍了几种现有运动参数估计方法的基本原理和不足。随后,基于投影误差最小判决函数,提出了用于双像运动估计的改进非线性迭代优化方法。数值仿真结果表明,在大平移小旋转角及小平移大旋转角2种运动条件下,采用文中提出的方法,运动估计精度均有所提高。此外,根据运动参数的估计值对真实目标进行三维重建实验,结果表明尺度重建误差小于2%且角度误差在3°以内。     

Recovery of ego-motion using region alignment

A method for computing the 3D camera motion (the ego-motion) in a static scene is described, where initially a detected 2D motion between two frames is used to align corresponding image regions. We prove that such a 2D registration removes all effects of camera rotation, even for those image regions that remain misaligned. The resulting residual parallax displacement field between the two region-aligned images is an epipolar field centered at the FOE (Focus-of-Expansion). The 3D camera translation is recovered from the epipolar field. The 3D camera rotation is recovered from the computed 3D translation and the detected 2D motion. The decomposition of image motion into a 2D parametric motion and residual epipolar parallax displacements avoids many of the inherent ambiguities and instabilities associated with decomposing the image motion into its rotational and translational components, and hence makes the computation of ego-motion or 3D structure estimation more robust     

Smoothed particle hydrodynamics: Applications to heat conduction

In this paper, we modify the numerical steps involved in a smoothed particle hydrodynamics (SPH) simulation. Specifically, the second order partial differential equation (PDE) is decomposed into two first order PDEs. Using the ghost particle method, consistent estimation of near-boundary corrections for system variables is also accomplished. Here, we focus on SPH equations for heat conduction to verify our numerical scheme. Each particle carries a physical entity (here, this entity is temperature) and transfers it to neighboring particles, thus exhibiting the mesh-less nature of the SPH framework, which is potentially applicable to complex geometries and nanoscale heat transfer. We demonstrate here only 1D and 2D simulations because 3D codes are as simple to generate as 1D codes in the SPH framework. Our methodology can be extended to systems where the governing equations are described by PDEs.     

三维人体姿态估计研究综述

三维人体姿态估计在本质上是一个分类问题和回归问题,主要通过图像估计人体的三维姿态。基于传统方法和深度学习方法的三维人体姿态估计是当前研究的主流方法。按照传统方法到深度学习方法的顺序对近年来三维人体姿态估计方法进行系统介绍,从而了解传统方法通过生成和判别等方法得到人体姿态的众多要素完成三维人体姿态的估计。基于深度学习的三维人体姿态估计方法主要通过构建神经网络,从图像特征中回归出人体姿态信息,大致可以分为基于直接回归方法、基于2D信息方法和基于混合方法的三维人体姿态估计这三类。最后对当前三维人体姿态估计研究所面临的困难与挑战进行阐述,并对未来的研究趋势做出展望。     

Maximal Discrepancy for Support Vector Machines

The Maximal Discrepancy (MD) is a powerful statistical method, which has been proposed for model selection and error estimation in classification problems. This approach is particularly attractive when dealing with small sample problems, since it avoids the use of a separate validation set. Unfortunately, the MD method requires a bounded loss function, which is usually avoided by most learning algorithms, including the Support Vector Machine (SVM), because it gives rise to a non-convex optimization problem. We derive in this work a new approach for rigorously applying the MD technique to the error estimation of the SVM and, at the same time, preserving the original SVM framework.     

Studies on the solution conformation and dynamics of a polysaccharide from Sinorhizobium fredii HH103 and its monosaccharide repeating unit

The conformational behavior of the homopolysaccharide isolated from Sinorhizobium fredii HH103 and its monosaccharide repeating unit (5-acetamido-3,5,7,9-tetradeoxy-7-(3-hydroxybutyramido)-L-glycero- L- manno-nonulosonic acid) was analyzed by nuclear magnetic resonance (NMR) spectroscopy and extensive molecular dynamics simulations (MD). The results indicate that the glycosidic linkages and lateral chains may adopt a variety of conformations. MD simulations using the generalized Born solvent-accessible surface area (GB/SA) continuum solvent model for water and the MM3* force field provide a population distribution of conformers that satisfactorily agrees with the experimental NMR data for the torsional degrees of freedom of the molecule.     

Dense Non-Rigid Motion Estimation in Sequences of Medical Images Using Differential Constraints

We describe a new method for computing the displacement vector field in time sequences of 2D or 3D images (4D data). The method is energy-minimizing on the space of correspondence functions; the energy is split into two terms, with one term matching differential singularities in the images, and the other constraining the regularity of the field. In order to reduce the computational time of the motion estimation, we use an adaptive image mesh, the resolution of which depends on the value of the gradient intensity. We solve numerically the minimization problem with the finite element method which gives a continuous approximation of the solution. We present experimental results on synthetic data and on medical images and we show how to use these results for analyzing cardiac deformations.