A mechano-geometric framework to characterize macromolecular ensembles

像蛋白质、RNA和DNA这样的大分子通过动态改变它们的三维结构来执行它们的细胞功能。了解这种分子的整体结构对于揭示其关键作用至关重要，并可能干预恢复失去的功能。虽然传统的分子动力学可以提供原子详细的轨迹，但它们的计算成本仍然巨大，通常限制了对小空间或时间尺度的分析。因此，较低的准确性，但高通量的方法是有价值的，可以获得第一个见解，并指导更详细的后续分析，集中在感兴趣的领域。机器人启发的运动学方法已经成功地应用于粗粒度分子建模：它们的高效性质允许快速，可靠地洞察分子运动，并为数据解释和集成提供有用的工具。在本提案中，我们的目标是将现有的运动学分子建模软件扩展到统一的机械几何框架中，以研究复杂大分子的构象集合。我们将把原子相互作用的几何约束公式与刚性理论结合起来，以获得对分子柔韧性的一阶见解。该公式将运动的内在层次强加到分子上，为在这种超高维环境中广泛而均匀地采样构象空间和能量景观提供了有效的工具。二面体自由度的大变化，强加的约束的违反，或通过空间或疏水接触的非天然相互作用可以报告对药物靶向重要的变构热点。通过研究几种蛋白质在其两种主要结构之间的转变，初步的令人兴奋的结果揭示了氨基酸的相互作用网络，这与已发表的实验数据非常一致。直接整合实验数据将有助于验证和改进我们的计算算法，并允许数据驱动的构象探索。由于实验数据往往稀疏且容易出错，我们的机械几何建模框架将有助于预测无法访问或未知的数据，并设计后续，更集中的实验。我们将重点研究传统和多温度晶体学、新一代x射线自由电子激光实验和双电子-电子共振（DEER）的实验数据。我们已经在几个系统（异氰化物水合酶（ICH），一种DJ-1超家族的酶，和g蛋白家族的两个成员，Gi和Gs）上获得了令人兴奋的数据，并对GPR126，一种黏附型g蛋白偶联受体，提供了令人满意的假设，为算法开发和优化提供了一个良好的开始。将这种方法与有限公司最优控制的基本经验相结合，我们的目标是揭示潜在的驱动力，这些驱动力可能会引导和稳定项目结束时的构象转变。

Macromolecules like protein, RNA, and DNA perform their cellular function through dynamically changing their three-dimensional structure. Understanding the overall structural ensemble of such a molecule is crucial to reveal its key roles, and potentially intervene to restore lost function. While traditional Molecular Dynamics can provide atomically detailed trajectories, their computational cost is still tremendous, often limiting analysis to small spatial or temporal scales. Hence, lower accuracy, but high throughput methods are valuable to obtain first insights and guide more detailed, subsequent analysis focused on the area of interest. Robotics-inspired kinematic methods have been applied with great success to coarse-grained molecular modeling: their efficient nature allows for fast, yet reliable insights into molecular motion, and provide useful tools for data interpretation and integration. In this proposal, we aim to extend our existing kinematic molecular modeling software into a unified, mechano-geometric framework to study conformational ensembles of complex macromolecules. We will combine geometric constraint formulations of atomic interactions with rigidity theory to gain first-order insights into molecular flexibility. This formulation imposes an intrinsic hierarchy of motions onto the molecule, providing an efficient tool to broadly and uniformly sample conformation space and energy landscape in such an ultra-high dimensional environment. Large changes of dihedral degrees of freedom, violation of imposed constraints, or non-native interactions via steric or hydrophobic contact can report on allosteric hotspots important for drug targeting. Initial exciting results from studying the transitions of several proteins between their two major configurations has revealed interaction networks of amino acids that were in great agreement with published experimental data. Directly integrating experimental data will help validate and improve our computational algorithm, and allow for data-driven conformational exploration. As experimental data is often sparse and prone to errors, our mechano-geometric modeling framework will help predict inaccessible or unknown data and design subsequent, more focused experiments.We will concentrate on experimental data from traditional and multi-temperature crystallography, new-generation X-ray free electron laser experiments and double electron-electron resonance (DEER). We already have exciting data on several systems (Isocyanide Hydratase (ICH), an enzyme of the DJ-1 hyperfamily, and two members of the G-protein family, Gi and Gs), and compelling hypotheses for GPR126, an adhesion-type G-protein coupled receptor, providing a head-start for algorithm development and optimization. Coupling this approach to the fundamental experience in optimal control at the LTD, we aim to reveal potential driving forces that may guide and stabilize conformational transitions towards the end of the project.