Framework for the Adaptive Multiscale Modeling of Biopolymers

提出的工作的主要目的是研究新的方法，为复杂的生物聚合物系统的行为的有效建模和仿真提供一种手段。这些系统通常具有在多个空间和时间尺度（水平）上发生的重要现象。细尺度（小而快速，例如单个原子的运动）现象对粗尺度（大而缓慢，例如基因表达在很大程度上取决于分子的形状（构象））行为的系统在大分子过程中很常见，对人类健康至关重要。为了真正深入了解重要细胞过程的行为和控制，人们必须了解它们背后的物理原理和机制。基于物理的建模和仿真将在获得这样的理解方面发挥关键作用。不幸的是，这些分子系统的计算成本如此之高，以至于目前无法将它们建模和模拟到足够的水平（在准确性和持续时间上）。由于这些分子过程的许多重要方面在感兴趣的过程中发生了重大变化，因此模型本身必须同样能够适应，以便它能够准确地表示重要过程，同时保持快速和成本效益。拟议的工作就是为此目的而进行的。这项工作涉及到生产自适应，多层次的建模策略，利用先进的多体方法。基于物理的内部度量将系统模型划分为区域，每个区域都有自己的局部时间和空间尺度集。区域边界和模型类型可能从原子（精细尺度）变化到连续（粗尺度），然后根据需要动态（自适应）调整，以最小的计算成本捕获重要的系统行为。潜在的FDCA和ODCA公式产生的方程本质上被划分为这些区域（子域）。此外，这些方程的总体结构是二叉树结构，因此所得公式非常有利于有效的并行计算机实现。这项工作的影响将大大增加这种复杂分子动力学系统的建模和分析的速度和程度。这将允许分析人员以比目前可能的更节省成本、时间和资源的方式来处理更复杂的系统，从而导致更好的理解。提出的适应性多尺度策略特别适用的这类系统的例子是包括RNA、DNA和蛋白质的生物聚合系统。所提出的框架有望提供一种更深入了解关键生物分子过程的方法，这可能有助于我们在未来学习修改和控制这些过程。这种理解和能力可以在许多积极方面对人类健康产生重大影响。

The principal objective of the proposed work is to research new methods, which provide a means for the efficient modeling and simulation of the behavior of complex bio-polymeric systems. These systems often have important phenomena taking place at multiple spatial and temporal scales (levels). Systems where fine scale (small and rapid, e.g. motions of individual atoms) phenomena contributes significantly to coarse scale (large and slow; e.g. gene expression depends to significant degree on the shape(conformation) of the molecule) behavior are common in macro-molecular processes and are essential to human health.To gain true insight into the behavior and control of important cellular processes, one must understand the physical principles and mechanisms that underlie them. Physics based modeling and simulation will play a critical role towards gaining such an understanding. Unfortunately, these molecular systems are so computationally costly that they cannot currently be modeled and simulated to an adequate level (in accuracy and duration). Because many of the important aspects of these molecular processes change significantly during the process of interest, the model itself must be similarly able to adapt so that it can accurately represent the important process, while remaining fast and cost effective.The proposed work is devoted to this end. This work involves to production of an adaptive, multi-level modeling strategy, utilizing advanced multibody methods. Physics-based internal metrics guide the division of the system model into regions, each with its own local temporal and spatial set of scales. The region boundaries and model types, which may vary from atomistic (fine scale) to continuum (coarsest scale), are then dynamically (adaptively) adjusted, as needed to capture important system behavior at minimum computational cost. The underlying FDCA and ODCA formulations produce equations which are inherently divided into such regions (subdomains). Additionally, the overall structure of these equations are those of a binary-tree, so the resulting formulation is highly conducive to effective parallel computer implementation. The impact of this work will be a great increase in the rate and extent to which such complex molecular dynamic systems may be modeled and analyzed. This will allow the analyst to treat far more complex systems in a more cost, time, and resource effective manner than is currently possible, thus leading to greater understanding. Examples of such systems where the proposed adaptive multiscale strategy should be particularly suitable are biopolymeric systems which include RNA, DNA, and proteins. The proposed framework is expected to provide a means to obtain greater insight into and understanding of key biomolecular processes, which may contribute greatly to our learning to modify and control such processes in the future. Such understanding and ability could significantly impact human health in many positive respects.