CCF: EAGER: Dimension Reduction and Optimization Methods for Flexible Refinement of Protein Docking

蛋白质是细胞中的“主力”分子。它们之间、核酸、配体和其他分子之间的相互作用是许多重要细胞过程的基础。因此，优先在原子水平上阐明蛋白质相互作用，对于理解这些过程并在患病细胞中治疗它们是重要的。蛋白质对接通过计算找到自由能最低的蛋白质构象来达到目的。由于两个原因，解决这样一个自由能量最小化问题是具有挑战性的。首先，搜索空间是非常高维的，因为蛋白质在相互作用时不是刚体，而是灵活的。其次，自由能函数的评估成本非常高，且难以优化。建议的研究直接解决了这些在细化阶段对蛋白质对接的挑战。首先，搜索空间的维数将大大降低。虽然蛋白质由数百到数千个原子组成，但并非所有这些原子的集体运动都具有物理意义。该方法采用新颖的正态模态分析工具，将原子集体运动空间分解为刚体运动空间和一些相关的柔性运动空间。其次，对能量最小值的搜索效率将大大提高。简化搜索空间中的能量景观将被表征，最先进的搜索方法将被应用于空间中的能量最小化。提出的研究还将对蛋白质相互作用的简化构象空间和简化空间中自由能函数的景观产生见解。更广泛的影响预测蛋白质如何高效和准确地相互作用的能力对于理解生命的结构和功能组织，开发诊断和治疗工具来治愈疾病，以及发现提高生物能源产量的方法具有巨大的价值。所提出的研究是跨学科的，因此有望对需要蛋白质对接工具来帮助他们研究的生物学家以及需要应用或开发降维和优化方法的计算机科学家有用。项目将为学生提供发展跨学科技能的培训机会。

Proteins are the 'workhorse' molecules in cells. Their interactions with each other, nucleic acids, ligands, and other molecules underlie many important cellular processes. Therefore, elucidating protein interactions, preferentially at the atomic level, is important to understanding these processes and treating them in diseased cells. Protein docking achieves the purpose computationally by finding the protein conformation of the lowest free energy values. Solving such a free energy minimization problem is challenging for two reasons. First, the search space is extremely high-dimensional because proteins are not rigid bodies but rather flexible when interacting. Secondly, the free energy function is very costly to evaluate and rugged to optimize. Intellectual MeritThe proposed research directly addresses these challenges to protein docking at the refinement stage. First, the dimension of the search space will be substantially lowered. Although proteins consist of hundreds to thousands of atoms, not all collective motions of those atoms are physically meaningful. The proposed methods will decompose the space of collective atomic motions into that of rigid-body motions and that of a few relevant flexibility motions by applying novel normal mode analysis (NMA) tools. Second, the search efficiency for the energy minimum will be substantially improved. Energy landscape in the reduced search space will be characterized and state-of-the-art search methods will be applied for energy minimization in the space. The proposed research will also produce insights on the reduced conformational space of protein interactions and the landscape of free energy functions in the reduced space. Broader ImpactsThe ability to predict how proteins interact with efficiency and accuracy is of tremendous value to understanding the structural and functional organization of life, developing diagnosis and therapeutics tools to cure diseases, and discovering approaches to improve bioenergy yields. The proposed research is interdisciplinary and thus expected to be useful to biologists who need protein docking tools to aid their study as well as computer scientists who need to apply or develop dimension reduction and optimization methods. Project will provide training opportunities to students in developing interdisciplinary skills.