Bayesian Statistics and Algorithms for Homology Modeling

描述(由申请人提供)：本次续展申请扩展了上一次资助期间的两个主题--用于结构确定和预测的蛋白质结构特征的现代统计分析，以及生物组装体结构的分析和预测。它还将我们的工作扩展到一个新的领域--对蛋白质晶体中生物相关相互作用的新观测进行实验测试。在第一个目标中，我们追求 初步观察到蛋白质侧链的Cβ原子的位置随旋转异构体以及主链二面体phi和psi的不同而不同。这种变化影响了整个侧链在结构确定和结构预测中的位置。我们将使用一种基于高斯过程回归的新配方，对每个chi1旋转体放置Cbeta所需的角度作为phi和psi的函数进行新的回归分析。这些将应用于侧链键角和二面角以及我们在上一次赠与期间调查的主链键角。这些主干和侧链回归函数将被添加到我们的程序SCWRL4和程序Rosetta中的侧链建模中。我们将进一步修改SCWRL4，以预测多个侧链位置及其占有率，用于配体对接和药物设计，并允许纳入核磁共振实验约束。我们的第二个主要目标是改进针对多个查询序列的生物组装的比较建模。这将包括改进整个PDB的家族和超家族分类，以及改进如何从蛋白质结构的不对称单位和晶体对称性推断生物组装。我们将开发一个用于生物组装的自动建模的服务器，该服务器将根据可用的模板提供几个预测的组装，并包括对查询和模板结构的大量生物注释。在一个新的方向上，我们将研究我们在可能是反式自磷酸化事件的快照的激酶晶体中观察到的同源二聚体相互作用。其中包括LCK和IGF1R的激活环交换二聚体，它们可能是几乎所有人类酪氨酸激酶的激活自动磷酸化事件的模型。将利用几种实验方法，并将对自动磷酸化界面进行进一步的结构分析。

DESCRIPTION (provided by applicant): This renewal application expands upon two themes of the previous grant period - modern statistical analysis of protein structural features for use in structure determination and prediction, and the analysis and prediction of the structures of biological assemblies. It also extends our work into a new area - experimental tests of novel observations of biologically relevant interactions in protein crystals. In the first aim, we pursue preliminary observations that the position of the Cbeta atom of protein side chains varies with rotamer and with the backbone dihedrals phi and psi. This variation affects the position of the entire side chain in structure determination and structure prediction. We will perform a new regression analysis of the angles required to place Cbeta for each chi1 rotamer as a function of phi and psi with a novel formulation based on Gaussian process regression. These will be applied to side chain bond angles and dihedral angles as well as backbone bond angles we have investigated in the previous grant period. These backbone and side- chain regression functions will be added to modeling side chains in our program SCWRL4 and in the program Rosetta. We will further modify SCWRL4 to perform prediction of multiple side-chain positions and their occupancies for use in ligand docking and drug design and to allow the incorporation of NMR experimental constraints. Our second major aim is to improve the comparative modeling of biological assemblies for multiple query sequences. This will involve improved family and superfamily classifications of the entire PDB as well as improvements in how biological assemblies are inferred from asymmetric units and crystal symmetry of protein structures. We will develop a server for automated modeling of biological assemblies that will provide several predicted assemblies based on the available templates and include substantial biological annotations for the query and the template structures. In a new direction, we will investigate our observation of homodimer interactions in crystals of kinases that may be snapshots of trans- autophosphorylation events. These include the activation-loop swapped dimers of LCK and IGF1R that may be models of the activation autophosphorylation events for nearly all human tyrosine kinases. Several experimental approaches will be utilized and further structural analysis of the autophosphorylation interfaces will be performed.