Development Of Theoretical Methods For Studying Biological Macromolecules

生物大分子研究理论方法的发展

• 批准号: 8158018
• 负责人: Bernard R Brooks
• 金额: $ 50.86万
• 依托单位: NATIONAL HEART, LUNG, AND BLOOD INSTITUTE
• 依托单位国家: 美国
• 资助国家: 美国
• 起止时间: 至
• 项目状态: 未结题
• 来源: https://reporter.nih.gov/project-details/8158018
• 关键词: Biological; Development; Methods; macromolecule

New theoretical techniques are being developed and characterized. These efforts are usually coupled with software development, and involve the systematic testing and evaluation of new ideas. This development is driven by current needs and interests. Self-guided molecular simulation for ensemble sampling Self-guided molecular dynamics (SGMD)/Langevin dynamics (SGLD) was developed for enhanced conformational search. This method makes rare events that otherwise not accessible by molecular dynamics simulation observable with modest computing resources. Typical applications include protein folding, signal transaction, water penetration, etc. A major challenge when applying SGMD/SGLD method in simulation studies is to quantitatively measure the deviation of the ensemble of states from a canonical ensemble and how to correct SGMD/SGLD results. This challenge has been a hurdle for the application of the method. By analyzing the characters of the guiding force and SGMD/SGLD simulation behavior, we derived a thermodynamic relation between a regular simulation and a self-guided simulation. By separating low frequency part from high frequency part, we are able to quantitative describe the enhancement in low frequency motion and correct by reweighing the deviation in the conformational distribution. This thermodynamic relation paves the way for many application and extension of the SGMD/SGLD method. One immediate extension is by introducing the relation into the equation of motion, we developed the so called SGMDfp/SGLDfp method that can produce directly a correct conformational distribution while enhances the conformational search using both force and momentum. Another application is to combine SGMD/SGLD with replica exchange with or without temperature differences. The advantage of SGMD/SGLD-replica exchange without temperature change is the reduction of the number of replicas needed for large systems such as solvated proteins. Multi-scale modeling Multi-scale modeling methods and application have become increasingly important for modeling important cellular events that are too complex to be studied at the rigorous QM level or even the all-atom classical force-field level. Usually, multi-scale techniques use high-level techniques to parameterize the coarser ones, however, it is also possible to couple these methods within a single framework to allow energies, forces, and NMA calculations to be done. MSCALE, a generalized approach for multi-scale modeling has therefore been developed and implemented within CHARMM. The MSCALE communication paradigm has also been made to work with other codes including Gaussian03, MOLPRO, NWCHEM, and the SANDER module of AMBER. The MSCALE communications protocol is easy to adopt for use with other programs. The following functionality exists within the CHARMM MSCALE implementation: 1. Triple-parallel scheme: (a) replicas, (b) Hamiltonian term or layer distribution, (c) parallelization of individual energy terms. 2. Multi-scale normal mode analysis with analytic Hessians. 3. Generalized approach to combining multiple physical and theoretical scales within a single calculation. For example, additive QM/MM, subtractive QM/MM, and all-atom/coarse-grained calculations are supported. 4. A general method for combining different multiscale approaches in a single calculation is supported. For example, both additive and subtractive QM/MM methods can be used in a single calculation. 5. Generalized free-energy perturbation calculations using multi-scale models. Hessian based methods for multi-scale modeling As part of the effort to develop generalized approaches to multi-scale simulations, new analysis techniques must be developed to interpret the results of multi-scale normal mode analysis calculations. Validating the results of these calculationss is a difficult problem because both the number of modes and frequencies and the lengths of the individual mode vectors differ from those produced by all-atom or more traditional coarse-grained NMA calculation (e.g. via the Elastic Network Model). Various approaches to this problems such as measuring spatial extents of modes or mapping lower level modes on a higher level representation are being studied and benchmarked. Excited states methods and applications The time-dependent density functional theory (TDDFT) is the method of choice for studying excited states in biological systems. In our recent benchmark calculations on the valence excited states in 20 small molecule, we found that TDDFT is more twice more accurate than configuration interaction singles (CIS), but only at a comparable computational cost. We also found that that TDDFT even produces more accurate excited state bond lengths than much more expensive correlation methods (such as CC2), while the adiabatic excitation energies are about the same quality. Recently, we achieved an efficient serial and parallel implementation of the analytical gradient for TDDFT (specificly its Tamm-Dancoff approximation). With that, we will study excited state dynamics of tryptophan in water and protein environments. Combining Conformational Space Annealing with Replica Exchange Method Replica exchange molecular dynamics (REMD) has been successfully used to improve the conformational search for model peptides and small proteins. However for larger and more complicated systems the use of REMD is still computationally intensive since the number of replicas required increases with system size. Achieving convergence with systems with slow transition kinetics is also very difficult. Several methods have been proposed to overcome the size and convergence speed issues of REMD. One of these methods is called Reservoir Replica Exchange Method (R-REMD) where the conformational search and temperature equilibration are separated. This allows integrating computationally efficient search algorithms with replica exchange. The Conformational Space Annealing (CSA) method has been shown to be able to determine the global free energy of proteins efficiently and has been used in structure prediction successfully. CSA uses a genetic algorithm approach to perform the conformational search and determine the minimum energy structure. We have used conformations generated through CSA method and collected them in a reservoir. Replica exchange was then performed where the top replica was seeded with the reservoir structures and fast convergence at every temperature is observed. Preliminary tests with model peptides show significant improvement on sampling efficiency and more thorough testing with small proteins is being done. Isotropic Periodic Sum Method for multipole interactions Multipole interactions play an important role in biological systems and increasing amount of effort have been contributed to include multipole interactions in force field development. The computation expense of multipole interactions are overwhelming for simulation. IPS method provides a convenient approach to tackle this problem. Base on IPS concept, we developed IPS potentials for multipole interactions. As a result, multipole interaction calculation can be carried out exactly as the cutoff method, with the IPS potentials to account for long range contributions. This development makes including multipole interactions a convenient way to extend force field developments and to speed up simulations with multipole interactions. Other ongoing method development projects: Improved conformational sampling by combining SGLD with replica exchange molecular dynamics QM/MM full Hessian and MBH Hessian Constant pH simulations with pH replica exchange Development and benchmarking of coarse-grained models in CHARMM Determination of test and evaluation criteria for sampling methods, CASA (Comparative Assessment of Sampling Algorithms)

新的理论技术正在开发和表征。 这些工作通常与软件开发相结合，并涉及对新想法的系统测试和评估。这种发展是由当前的需求和利益驱动的。 用于整体采样的自引导分子模拟 自引导分子动力学 (SGMD)/朗之万动力学 (SGLD) 是为了增强构象搜索而开发的。 这种方法使得分子动力学模拟无法通过适度的计算资源观察到的罕见事件成为可能。 典型应用包括蛋白质折叠、信号处理、水渗透等。 在仿真研究中应用SGMD/SGLD方法时的一个主要挑战是定量测量状态系综与规范系综的偏差以及如何纠正SGMD/SGLD结果。 这一挑战一直是该方法应用的障碍。 通过分析引导力的特点和SGMD/SGLD模拟行为，我们推导了常规模拟和自引导模拟之间的热力学关系。通过将低频部分与高频部分分开，我们能够定量描述低频运动的增强，并通过重新权衡构象分布的偏差进行校正。 这种热力学关系为 SGMD/SGLD 方法的许多应用和扩展铺平了道路。 一个直接的扩展是通过将关系引入到运动方程中，我们开发了所谓的 SGMDfp/SGLDfp 方法，该方法可以直接产生正确的构象分布，同时使用力和动量增强构象搜索。 另一个应用是将 SGMD/SGLD 与有或没有温差的副本交换结合起来。 在不改变温度的情况下进行 SGMD/SGLD 复制品交换的优点是减少了大型系统（例如溶剂化蛋白质）所需的复制品数量。 多尺度建模 多尺度建模方法和应用对于建模重要的细胞事件变得越来越重要，这些事件过于复杂，无法在严格的量子力学水平甚至全原子经典力场水平上进行研究。通常，多尺度技术使用高级技术来参数化较粗略的技术，但是，也可以将这些方法耦合在单个框架内，以允许完成能量、力和 NMA 计算。因此，在 CHARMM 中开发并实施了 MSCALE，一种多尺度建模的通用方法。 MSCALE 通信范例还可以与其他代码一起使用，包括 Gaussian03、MOLPRO、NWCHEM 和 AMBER 的 SANDER 模块。 MSCALE 通信协议很容易与其他程序一起使用。 CHARMM MSCALE 实现中存在以下功能： 1. 三重并行方案：(a) 副本，(b) 哈密顿项或层分布，(c) 各个能量项的并行化。 2. 使用解析 Hessian 进行多尺度正态模态分析。 3. 在一次计算中结合多个物理和理论尺度的通用方法。例如，支持加法QM/MM、减法QM/MM以及全原子/粗粒度计算。 4. 支持在单个计算中组合不同多尺度方法的通用方法。例如，加法和减法 QM/MM 方法都可以在单个计算中使用。 5. 使用多尺度模型的广义自由能扰动计算。 基于 Hessian 的多尺度建模方法 作为开发多尺度模拟通用方法的一部分，必须开发新的分析技术来解释多尺度简正模态分析计算的结果。 验证这些计算的结果是一个困难的问题，因为模式和频率的数量以及各个模式向量的长度都不同于全原子或更传统的粗粒度 NMA 计算（例如通过弹性网络模型）产生的结果。 正在研究和基准测试解决此问题的各种方法，例如测量模式的空间范围或将较低级别的模式映射到较高级别的表示上。 激发态方法和应用 瞬态密度泛函理论 (TDDFT) 是研究生物系统中激发态的首选方法。在我们最近对 20 个小分子的价激发态进行的基准计算中，我们发现 TDDFT 的准确度比构型相互作用单原子 (CIS) 高出两倍，但前提是计算成本相当。我们还发现，TDDFT 甚至比更昂贵的相关方法（例如 CC2）产生更准确的激发态键长，而绝热激发能的质量大致相同。最近，我们实现了 TDDFT 解析梯度的高效串行和并行实现（特别是其 Tamm-Dancoff 近似）。这样，我们将研究水和蛋白质环境中色氨酸的激发态动力学。 构象空间退火与副本交换方法相结合 复制交换分子动力学 (REMD) 已成功用于改进模型肽和小蛋白质的构象搜索。然而，对于更大、更复杂的系统，REMD 的使用仍然是计算密集型的，因为所需的副本数量随着系统规模的增加而增加。实现具有慢转变动力学的系统的收敛也非常困难。已经提出了几种方法来克服 REMD 的尺寸和收敛速度问题。其中一种方法称为储库复制品交换法 (R-REMD)，其中构象搜索和温度平衡是分开的。这允许将计算效率高的搜索算法与副本交换集成起来。构象空间退火（CSA）方法已被证明能够有效地确定蛋白质的整体自由能，并已成功用于结构预测。 CSA 使用遗传算法方法来执行构象搜索并确定最小能量结构。我们使用通过 CSA 方法生成的构象并将其收集在储库中。然后进行副本交换，其中顶部副本被植入储层结构，并观察到在每个温度下的快速收敛。模型肽的初步测试显示采样效率显着提高，并且正在对小蛋白质进行更彻底的测试。 多极相互作用的各向同性周期和法 多极相互作用在生物系统中发挥着重要作用，并且越来越多的努力致力于将多极相互作用纳入力场开发中。 多极相互作用的计算费用对于模拟来说是巨大的。 IPS方法为解决这个问题提供了一种便捷的方法。 基于IPS概念，我们开发了多极相互作用的IPS潜力。 因此，多极相互作用计算可以完全按照截断法进行，并且 IPS 潜力可以考虑长程贡献。 这一发展使得多极相互作用成为扩展力场发展并加速多极相互作用模拟的便捷方法。 其他正在进行的方法开发项目： 通过将 SGLD 与复制品交换分子动力学相结合改进构象采样 QM/MM 完整 Hessian 和 MBH Hessian 通过 pH 副本交换进行恒定 pH 模拟 CHARMM 中粗粒度模型的开发和基准测试 采样方法测试和评估标准的确定，CASA（采样算法的比较评估）