Developing Mechanisms from Protein Structures

在一组研究中，我们一直在研究蛋白质中最保守的部分，这些部分占总结构的一小部分，通常为6 - 20个残基。同时，我们也在研究各种分子性质，如包装，与序列保守区的相关性。了解序列保守性结构的物理基础将使我们能够仅从序列中对蛋白质核心进行一些关键性的预测。我们研究了几种蛋白质，最近的溶菌酶和α-乳白蛋白，在那里我们发现了一个核的残基连接在所有的二级结构元素，并可以作为一个关键的核。此外，我们正在开发几种新的线程方法，以全面考虑核心和序列保护。我们也一直在研究蛋白质折叠中间体的荧光，以确定分子内的距离，以比较这些与天然结构的距离。在迄今为止的情况下，在熔融球中间体的距离是类似的天然状态的距离。 计算生物学的另一个目标是理解分子机制。用传统的分子动力学方法处理蛋白质结构并不能提供大尺度运动的信息。我们正在研究蛋白质动力学与一个新的粗粒度模型，每个残基只有一个点。这种新方法代表了一种从结构中推断功能行为的简单方法。它考虑了基于高斯网络模型的已知蛋白质结构的波动。这个程序已被证明是令人满意的样品周围的天然构象蛋白质的残基波动的分布，并产生良好的协议与晶体温度因子和氢交换数据，为各种各样的蛋白质和核酸结构。由于该方法简单，结果直观且令人信服。该方法产生一系列的运动模式，通常是铰链弯曲运动，甚至包括最慢的，最全局的运动。在最近的方法开发中，我们已经将计算从标量扩展到矢量，因此我们现在能够跟踪平移变形。这为理解超大甚至超分子结构的总功能动力学开辟了新的和令人兴奋的前景。在另一个计算改进中，计算所需的时间已经显著减少了几个数量级。我们最近的研究包括：1）逆转录酶，其中我们展示了手指/拇指结合位点和核糖核酸酶H位点的运动之间的反相关性如何导致核酸链通过酶以释放-拉动-转动系列运动的逐步加工机制; 2）游离的t-RNA并与其同源合成酶结合（两者都表现出类似的运动，独立和一起）; 3）GroEl-GroES蛋白伴侣系统，其是非常大的系统（8000个残基），以显示腔如何以许多不同的方式被压缩以及可用的结合内部结合表面如何通过这些运动而改变;和4）微管蛋白，其中二聚化对于增强运动的协同性是关键的，并且二聚体运动包括亚基之间的摆动，沿二聚体长轴的沿着伸长和压缩，以及两种单体在彼此相反的方向上的扭转。最近的应用包括微管蛋白的纤维形式，以及其他几种核酸结合蛋白。预期的应用包括研究各种蛋白质的结合和构象转变。 我们现在可以将这些计算扩展到非常大的结构（100万个氨基酸）。功能运动，例如加工，通常只取决于形状，而不是结构的所有细节。通常这些可以是两个结构域之间的铰链，整个结构的拉伸，亚基之间的旋转。重要的发现是，在这些大尺度运动中，结合位点上的一些局部环打开和关闭，而不是独立的。这对序列保守性具有重要意义。整个蛋白质结构对于促进这些明显的局部运动至关重要! 新的快速结构测定。通过结合计算的结构库与表面距离测量使用交联结构，其链接由质谱鉴定。 磷酸化调节蛋白质。级联效应与构象的变化有关。我们开始了解为什么磷酸化会引起构象变化，或者不会。 药物发现。同样的数学形式主义（奇异值分解）用于计算蛋白质的运动的应用程序也已经分析了细胞系筛选数据。可以将122种药物分成25个不同的组，并以高度系统化的方式对细胞系本身进行分类。60种细胞系分为21组，其中最强的分组是肾癌、白血病和卵巢癌。 Z01 BC 08370 - 17

In one group of studies we have been studying the very most conserved parts of proteins, which comprise a small fraction of the total structure, typically 6 - 20 residues. At the same time we are also investigating the correlations of various molecular properties, such as packing, with the sequence conserved regions. Understanding the physical basis in structure for sequence conservation would enable us to make some critical predictions of protein cores from sequences alone. We studied several proteins, most recently lysozyme and a-lactalbumin, where we find a nucleus of residues that connect among all of the secondary structure elements and could act as a critical nucleus. In addition we are developing several new approaches to threading to consider cores and sequence conservation in a comprehensive way. We also have been studying protein folding intermediates by fluorescence to determine intramolecular distances in order to compare these with the native structure distances. In cases to date, the distances in the molten globule intermediates are similar to the native state distances. Another goal of computational biology is to understand molecular mechanisms. Protein structures treated with conventional molecular dynamics have not been so informative about large scale motions. We are investigating protein dynamics with a new coarse-grained model having only one point per residue. This new approach represents a simple way to infer functional behavior from structures. It considers fluctuations about known protein structures based on a Gaussian network model. This procedure has been shown to sample satisfactorily the distribution of residue fluctuations around the native conformation in proteins, and to yield excellent agreement with crystallographic temperature factors and hydrogen exchange data, for a broad variety of proteins and nucleic acid structures. Since this method is simple, results are intuitive and compelling. The approach yields a series of modes of motion, typically hinge bending motions, including even the slowest, most global motions. In a recent method development we have extended the calculations from scalar to vector, so that we are now able to follow translational deformations. This opens new and exciting prospects for comprehending the total functional dynamics of extremely large, even supra-molecular structures. In another computational improvement, the time required for calculations has been significantly reduced by several orders of magnitude. Our recent studies with this approach have included: 1) reverse transcriptase in which we showed how the anti-correlations between the motions of the fingers/thumb binding site and the ribonuclease H site could lead to a step-wise processing mechanism for the progression of the nucleic acid chain through the enzyme in a release-pull-turn series of motions; 2) t-RNA free and bound to its cognate synthetase (both show similar motions, independently and together); 3) the GroEl-GroES protein chaperone system which is an extremely large system (8000 residues), to show how the cavity is compressed in many different ways and how the available binding interior biding surface changes through these motions; and 4) tubulin where the dimerization is critical for enhancing the cooperativity of motions and the dimer motions include a wobble between the subunits, elongation and compression along the long axis of the dimer, and twisting of the two monomers in directions opposite to one another. Recent applications have included tubulin in its fibrillar form, as well as several other nucleic acid binding proteins. Anticipated applications include studies of binding and conformational transitions for a broad variety of proteins. We now can extend these calculations to extremely large structures (1 million amino acids). Functional motions, e.g. processing, usually depend only on shape and not on all details of structure. Typically these can be hinges between two domains, stretching of whole structure, rotations between subunits. The important finding is that some local loops over binding sites open and close during some of these large scale motions, not independently. This has important implications for sequence conservation. The entire protein structure is critical for facilitating these apparently local motions! New Fast Structure Determination. By combining computed structure libraries with surface distance measurements using crosslinked structures whose links are identified by mass spectroscopy. Phosphorylation to Regulate Proteins. Cascade effects relate to changes in conformation. We are beginning to develop some understanding of why phosphorylation either causes conformational change or it does not. Drug Discovery. An application of the same mathematical formalism (singular value decomposition) utilized for calculating the motions of proteins has also been made to analyze the cell-line screening data. It was possible to cluster the 122 agents into 25 distinct groups, as well as to classify the cell lines themselves, in a highly systematic way. The 60 cell lines cluster into 21 groups, with the strongest groupings found for renal, leukemia and ovarian cancer. Z01 BC 08370-17