CDI-Type I Collaborative Research: Multi-Scale Modeling of Protein-Modulated DNA Large-Scale Dynamics by Free Energy Surface Matching

蛋白质-DNA的相互作用基本上控制着细胞内所有主要的遗传交易，例如DNA复制、修复、转录和重组。这些作用从蛋白质- dna结合的纳米级位点局部开始，通常会产生非常大的，？Ým-scale DNA构象变化。在这些复杂的动力系统中，极其广泛的长度和时间尺度给计算建模带来了巨大的挑战。该项目通过提出一种变革性的多尺度计算方法来解决这一挑战，该方法可以在生物学相关的长度/时间尺度（例如，微米/毫秒或更长）上捕获大型蛋白质- dna复合物的时间演化。我们的方法从（一次性）大规模并行MD计算和蛋白质结构域的伞式采样开始，在DNA双链的短片段（原子细节）存在下建立一个不受干扰的自由能表面。接下来，我们将长DNA双链的棒状模型耦合到蛋白质上（通过几何蛋白质边界条件），并形成整个蛋白质-DNA复合物的主要自由度的降阶动力系统（Fokker-Planck概率）模型。由此产生的Fokker-Plank模型捕获了杆/DNA和蛋白质结构域的完整（双向）动态耦合，并能够在所需的长长度/时间尺度上进行整合。我们在两个大型且具有挑战性的系统上演示了我们的方法；即1)人类拓扑异构酶I对DNA超级线圈的松弛，以及2)噬菌体中病毒衣壳上dsDNA的包装和喷射。本研究旨在解决长期存在的挑战，即在长时间尺度上预测大型生物分子系统的动力学行为。这些预测对于理解基本的细胞过程（包括DNA转录、复制和修复）和解释单分子实验的令人兴奋的结果是必不可少的。更广泛地说，我们的方法提供了一种系统的方法，可以在广泛的领域中将原子级和连续统级（例如微米级）的物质描述结合起来。其他领域可能包括DNA/RNA复合物，形成用于脚手架，计算或纳米推进的大规模核酸（折纸）结构；碳纳米管与有机和无机纳米颗粒的相互作用纳米线及其在生物分子检测中的应用以及鞭毛、胶原纤维和细胞骨架元素（如肌动蛋白、神经丝、微管）等的结构动力学。

Protein-DNA interactions govern essentially all major genetic transactions within the cell including, for example, DNA replication, repair, transcription and recombination. These actions, which begin locally at the nm-sized site of protein-DNA binding, often generate very large, ?Ým-scale DNA conformational changes. The enormously broad length and time scales invoked in these complex dynamical systems create formidable challenges for computational modeling. This project addresses this challenge by proposing a transformative, multi-scale computational method that captures the time-evolution of large protein-DNA complexes on biologically relevant length/time scales (e.g., micron/millisecond and longer). Our method begins with (one-time) massively parallel MD computations and umbrella sampling of the protein domain to establish an unperturbed free energy surface in the presence of a short (atomically-detailed) fragment of the DNA duplex. Next, we couple a rod model of the long DNA duplex to the protein (by geometric protein boundary conditions) and form a reduced-order dynamical system (Fokker-Planck probability) model of the entire protein-DNA complex for the dominant degrees of freedom. The resulting Fokker-Plank model captures the complete (two-way) dynamic coupling of the rod/DNA and protein domains and enables integration over the desired long length/time scales. We illustrate our method on two large and challenging systems; namely 1) the relaxation of DNA supercoils by human topoisomerase I, and 2) the packing and ejection of dsDNA from viral capsids in bacteriophages.This research aims at long standing challenges in predicting the dynamical behavior of large biomolecular systems on long length and time scales. These predictions are essential for understanding fundamental cellular processes (including DNA transcription, replication, and repair) and interpreting exciting results from single molecule experiments. More broadly, our method provides a systematic means to couple atomistic to continuum level (e.g. micron-scale) descriptions of matter in a wide range of fields. Other fields may include DNA/RNA complexes that form large scale nucleic acid (origami) structures for scaffolding, computing, or nanopropelling; carbon nanotubes interacting with organic and inorganic nanoparticles; nanowires and their use for bio-molecular detection; and the structural dynamics of flagella, collagen fibers and cellular cytoskeleton elements (e.g., actin, neurofilaments, microtubules), among others.