Continual Development of Protein Docking Algorithms

项目描述（由申请人提供）：本项目的目标是开发预测蛋白质-蛋白质复合物三维结构的计算方法、算法和软件。这些方法对于深入了解相互作用蛋白质的生物学功能和指导新型药物的设计是非常有价值的。这个项目以我们实验室现有的优势为基础。我们构建了几种蛋白质对接算法（ZDOCK， RDOCK和ZRANK），一个对接基准和一个对接服务器，被其他研究人员广泛使用。我们最近开发了一个新的能量功能IFACE，它提高了ZDOCK的性能。我们在将ZRANK与侧链构象搜索相结合方面取得了重大进展。该项目的前两个目标涉及算法开发。在Aim 1中，我们建议通过使用不同类型的蛋白质-蛋白质复合物进行训练来进一步开发IFACE潜力，并允许潜力依赖于距离。我们还计划开发一种五维快速傅立叶变换算法ZDOCK5D，用于有效地采样两个刚性蛋白质之间的六个刚体自由度。这项工作将利用我们为ZDOCK开发的评分专业知识，并将产生显着的速度改进。还计划开发一种通用的集中搜索方法，这将提高准确性和速度。在目标2中，我们将开发明确探索侧链灵活性的方法。我们将预测哪些残基在络合物形成时可能发生侧链构象变化。我们计划将ZRANK与RosettaDock（由Gray和Baker实验室开发的对接程序）结合起来执行侧链搜索。我们将进一步开发ZRANK，使其能够有效地重新排序那些侧链、旋转和平移位置已经被RosettaDock改进的结构。目标3的重点是开发一个管道来更新我们的对接基准，并在这个项目的4年过程中产生两次更新。我们还将进一步开发基于原子接触向量的分类器，用于区分瞬态配合物和专性配合物，这是更新基准和开发对接算法目标函数的重要一步。在目标4中，我们描述了改进现有计算套件的软件工程的计划。这项工作将有利于未来的算法和软件开发，符合目标1-3中概述的建议。我们将进一步开发我们的对接服务器。服务器上可用的软件将被扩展，以包含该应用程序中计划的许多新功能，服务器硬件将被分配更多的计算能力。最后，我们将加强我们的用户支持，以最大限度地发挥我们的程序套件对用户社区的影响。公共卫生相关性：蛋白质复合物的三维（3D）结构的知识提供了对组成蛋白质的生物学功能的见解，并有助于蛋白质药物的设计。我们的实验室已经开发了一套广泛使用的蛋白质-蛋白质对接算法和一个广泛使用的对接基准。该项目的目标是继续开发这些计算方法、算法和软件来研究蛋白质-蛋白质相互作用，以便在蛋白质对接社区中得到更广泛的应用。

DESCRIPTION (provided by applicant): The goal of this project is to develop computational methods, algorithms and software for predicting the three dimensional structures of protein-protein complexes. Such methods are extremely valuable for providing insights into the biological functions of the interacting proteins and can guide the design of novel pharmaceuticals. This project builds on the existing strengths in our lab. We have built several protein docking algorithms (ZDOCK, RDOCK and ZRANK), a docking benchmark and a docking server, which are widely used by other researchers. We have recently developed a new energy function, IFACE, which improves the performance of ZDOCK. We have made significant progress on combining ZRANK with side-chain conformation search. The first two Aims of the project regard algorithm development. In Aim 1, we propose to further develop the IFACE potential by using different types of protein-protein complexes for training, and allowing the potential to be distance-dependent. We also plan to develop a five-dimensional fast Fourier transform algorithm ZDOCK5D for efficiently sampling the six rigid-body degrees of freedom between two rigid proteins. This effort will take advantage of the scoring expertise we have developed for ZDOCK, and will yield significant speed improvements. The development of a general method for focused searching is also planned, which will improve both accuracy and speed. In Aim 2, we will develop methods to explicitly explore side chain flexibility. We will predict which residues are likely to undergo side chain conformational change upon complex formation. We plan to combine our ZRANK with RosettaDock (a docking program developed by the labs of Gray and Baker) to perform side chain searches. We will further develop ZRANK so that it can effectively rerank the structures for which side-chains and rotational and translational placements have been refined by RosettaDock. Aim 3 is focused on developing a pipeline for updating our docking benchmark and producing two updates during the 4-year course of this project. We will also further develop classifiers based on atomic contact vectors for distinguishing transient complexes from obligate complexes, an important step for both updating the benchmark and for developing target functions of docking algorithms. In Aim 4, we describe plans to improve the software engineering of our existing computational suite. This undertaking will benefit future algorithmic and software development, in line with the proposals outlined in Aims 1-3. We will further develop our docking server. The software available on the server will be extended to include many new features planned in this application and the server hardware will be allotted more computing power. Finally, we will strengthen our user support to maximize the impact of our suite of programs on the user community. PUBLIC HEALTH RELEVANCE: Knowledge of the 3-dimensional (3D) structures of protein complexes provides insights into the biological functions of the component proteins and can aid the design of protein drugs. Our lab has developed a suite of widely used protein-protein docking algorithms and a widely used docking benchmark. The goal of this project is to continue to develop these computational methods, algorithms, and software for studying protein-protein interactions, in order to foster broader use in the protein docking community.