Fundamentals of Ligand-Protein Interactions

配体-蛋白质复合物的两个配偶体的构象变化，即小分子配体在其蛋白质结合位点（在许多情况下是酶的催化活性位点）的构象变化是许多药物作用的中心方面，也是药物设计的计算方法中的关键挑战。在这一领域最早的出版物之一中，我们发现，对于蛋白质数据库（PDB）和剑桥结构数据库（CSD）中出现的一小部分配体，柔性化合物通常不会以其整体真空能量构象与蛋白质结合，并且通常甚至不会以任何局部真空能量构象结合。虽然这项研究使用了当时最大的数据集和最好的方法，但无论是实验数据库中的结构数量还是可用的软件和硬件资源都呈指数级增长。因此，我们正在重新审视这个重要的主题，分析数量级更多的结构，并在高水平的计算量子化学理论进行计算。在该项目迄今为止取得的其他里程碑中，我们提取了PDB的LigandExpo中最近提供的所有小分子配体。截至2008年5月，这是一组超过350，000个不同的3D坐标集（当前集大小：190万）。我们添加了来自几个不同来源的广泛注释。在一系列过滤器中使用这些注释，我们已经生成了高质量和可靠性的配体结构的“高质量”子集，根据所应用的严格性，其数量从大约1000次到大约5,000次。我们已经进行了高层次的量子化学计算这些高品质的配体集的构象能。在第一轮中，真空能量计算部分在我们自己的Linux集群上运行，部分在CIT，NIH的Biowulf集群上运行。在这个计算量巨大的项目中，同时使用了多达1000个CPU，单个作业需要几个小时到几个星期的CPU时间。我们从成功完成的约360次运行中获得了结果。这些结果清楚地表明，高构象能量的可能性完全证实了这些量子化学计算。为了探索水环境对配体构象能的可能影响-毕竟，真空并不是药物分子通常工作的地方-使用Gaussian 03中的SCI-PCM溶剂模型进行了第二轮量子化学计算。这些运行在计算机资源方面的要求甚至比真空计算更高。为了分析作为位置不确定性的函数的能量不确定性，这反过来又是晶体学分辨率的函数，我们进行了采样的构象与分辨率相关的扭转分布集中在分子力学力场水平的晶体结构构象。自由电子激光器等更强大的实验仪器的出现为回答这些问题开辟了新的可能性。与这个主题相关的是最近开始的一项关于有机小分子互变异构的研究，这是化学信息学和数据库中的一个重要问题（项目3），有效的药物设计（项目2和3），以及目前更好地理解蛋白质-配体相互作用和晶体结构的项目。这项工作由Laura Guasch-Pamies博士完成。互变异构的工作正在继续与有趣的结果相结合的实验，量子化学和化学信息学分析，已被描述的论文中接受或正在审查，和/或研究进一步的分析。这一领域的工作为涉及醛糖还原酶和其他癌症和HPV相关蛋白质的晶体结构的令人兴奋的新研究奠定了基础。在这个项目中发表了一些回顾该领域现状和未来可能性的论文，包括最近超高分辨率晶体学的作用。随着CADD工作组对对接工作的日益重视，这一主题具有了新的重要性。

The conformational changes of both partners of a ligand-protein complex, the small-molecule ligand in its the protein binding site (in many cases the catalytically active site of an enzyme) are a central aspect many drug actions, as well as a crucial challenge in computational approaches to drug design. In one of the earliest publications in the this field, we showed for a small set of ligands occurring both in the Protein Data Bank (PDB) and the Cambridge Structural Database (CSD) that flexible compounds are not usually bound to a protein in their global vacuum energy conformation, and oftentimes not even in any local vacuum energy conformation. While this study used the largest set of data and best methodology available at that time, both the number of structures in either experimental database and the software and hardware resource available have since grown exponentially. We are thus revisiting this important topic with an analysis of orders of magnitudes more structures, and computations performed at a high level of computational quantum-chemical theory. Among other milestones achieved so far in this project, we have extracted all occurrences of small-molecule ligands recently made available in PDB's LigandExpo. As of May 2008, this was a set of over 350,000 distinct sets of 3D coordinates (current set size: 1.9 million). We have added extensive annotation coming from several different sources. Using these annotations in a chain of filters, we have generated "high-quality" subsets of ligand structures of high quality and reliability numbering from just about one thousand to about 5,000 occurrences depending on the stringency applied. We have conducted high-level quantum-chemical calculations of conformational energies for these high-quality ligand sets. In the first round, vacuum energy calculations were run partly on our own Linux cluster, partly on the Biowulf cluster of the CIT, NIH. Up to a thousand CPUs were used simultaneously in this computationally massive project, with individual jobs taking from a few hours to several weeks of CPU-time. We obtained results from about 360 runs that completed successfully. These results clearly showed that the possibility for high conformational energies are fully confirmed by these quantum-chemical calculations. To explore the possible influence of aqueous environment on ligand conformational energies - after all, vacuum is not really where drug molecules typically operate - a second round of quantum chemical calculations was conducted, employing the SCI-PCM solvent model in Gaussian 03. These runs were even more demanding in terms of computer resources than the vacuum calculations. To analyze the energetic uncertainty as a function of the positional uncertainty, which in turn is a function of the crystallographic resolution, we conducted sampling of conformations with a resolution-dependent torsion distribution centered around the crystal structure conformation at the molecular mechanics force field level. The advent of ever more-powerful experimental instrumentation such as free electron lasers opens up new possibilities in answering these questions. Related to this topic is a study recently begun on tautomerism of small organic molecules, which is an important question both in chemoinformatics and databases (Project 3), efficient drug design (Projects 2 and 3), and the present project of better understanding protein-ligand interactions and the crystal structures aiding in this quest. This work has been performed by Dr. Laura Guasch-Pamies. The tautomerism work is continuing with interesting results of combined experimental, quantum-chemical and chemoinformatics analysis, which have been described in papers either accepted or under review, and/or studied for further analyses. The work in this field has laid the groundwork for exciting new research involving crystal structures of aldose reductase and other cancer- and HPV-related proteins. A number of papers reviewing the current status and the future possibilities of the field have been published in this project, including most recently the role of ultra-high resolution crystallography. This topic has gained new importance with the increased emphasis of docking in the CADD Group's work.