High Resolution in Single Particle Reconstruction

描述（由申请人提供）：在此更新申请中，重点将放在开发稳健的单颗粒低温电镜分析方法上，这些方法从一开始就包含对结果的统计验证。我们将集中在三个具体领域：(1)二维图像比对；(2)从头计算结构确定；(3)基于数据重采样方法的大分子构象变异性定量描述。在(1)中，一种基于最大似然（ML）范式的新型二维图像对齐方法将利用期望最大化算法并结合精确的图像形成模型。该方法的可行性需要出色的计算效率，这将通过在有限集合上跨越似然函数中的角参数来实现，符合典型电磁数据的特性。我们将能够通过在傅里叶谐波基础上使用滤波来优化所提出方法的性能，以模拟图像数据的对齐引起的模糊。我们还建议使用从ML中获得的对齐参数协方差矩阵的特征分析来直接从对齐信息中对图像进行分类。在(2)中，我们将通过引入考虑相交傅立叶平面之间的二维重叠之后的附加差异项，大大提高先前开发的基于公共线的方法的性能和可靠性。结合改进的对准算法，该方法将产生一种强大的方法来生成初始低温电镜结构，该方法将克服当前由于低信噪比和数据结构异质性而造成的限制。在(3)中，一种新的数据重采样方法旨在克服低温电镜数据集中投影的强各向异性分布所带来的限制，将允许对三维重建进行方差和协方差估计的自动化。通过数据重采样生成的大体积集的特征分析将用于（直接从图像数据）计算描述大分子组装的构象模式的特征向量。通过本特征向量分析确定的数据子集计算出的结构将为分子功能研究提供更高分辨率的构象模型。我们提出的方法不是渐进式的改进，而是代表了解决当前阻碍单粒子低温电镜进一步发展的具体问题的新方法。为了确保最大的可移植性和有效的传播，这些新方法将在目前部署的SPARX图像处理包中实现。公共卫生相关性：高分辨率冷冻电子显微镜（cryo-EM）已成为测定大型大分子复合物结构/功能的重要工具。即使在亚纳米分辨率下，低温电镜图也提供了丰富的结构信息，最终导致二级结构的确定，正如我们对核糖体结构的研究所证明的那样。此外，低温电镜是一种独特的结构技术，它能够检测一个样品中大分子组装的构象变异性，该样品可能包含各种构象状态的复合物的混合物。我们建议开发专用的数据处理和统计工具，以建立数据集中的构象数量，并用于研究结构的构象模式，直接从EM数据中获得。

DESCRIPTION (provided by applicant): In this renewal application, the focus will be on the development of robust single particle cryo-EM analysis methods that incorporate, from their inception, statistical verification of the results. We will concentrate on three specific areas: (1) 2-D image alignment, (2) ab initio structure determination, and (3) quantitative description of macromolecular conformational variability based on data resampling methodology. In (1), a novel 2-D image alignment approach based on the Maximum Likelihood (ML) paradigm will make use of an Expectation Maximization algorithm and incorporate a precise image formation model. The feasibility of the method requires excellent computational efficiency, which will be achieved by spanning the angular parameters in the likelihood function over a finite set, in agreement with properties of typical EM data. We will be able to optimize performance of the proposed method by using filtration in the Fourier Harmonics basis to model alignment-induced blurring of the image data. We also propose to use the eigenanalysis of the alignment parameter covariance matrix obtained from ML to classify images directly from alignment information. In (2), we will greatly improve the performance and reliability of previously developed common lines-based methodology by introducing additional discrepancy terms that follow from considering 2-D overlap between intersecting Fourier planes. In combination with improved alignment algorithms, this methodology will result in a robust approach for generation of initial cryo-EM structures that will overcome current limitations due to low Signal-to-Noise Ratio and structural heterogeneity of the data. In (3), a novel data resampling approach designed to overcome limitations arising from the strongly anisotropic distribution of projections in cryo-EM data sets will permit automation of variance and covariance estimation for 3-D reconstructions. Eigenanalysis of large volume sets generated through data resampling will be used to calculate (directly from the image data) eigenvectors describing the conformational modes of a macromolecular assembly. Structures calculated from data subsets identified through this eigenvector analysis will provide higher-resolution models of conformations relevant for studies of molecular function. Rather than incremental improvements, the methods we propose to develop represent novel approaches that address specific issues currently hindering further development of single particle cryo-EM. To assure maximum portability and efficient dissemination, these new methods will be implemented within the currently deployed SPARX image processing package. PUBLIC HEALTH RELEVANCE: High-resolution cryo-electron microscopy (cryo-EM) has become an important tool for the structure/function determination of large macromolecular complexes. Even at subnanometer resolution cryo-EM maps provide a wealth of structural information, eventually leading to determination of the secondary structure, as demonstrated by our work on the structure of the ribosome. In addition, cryo-EM is a unique structural technique in its ability to detect conformational variability of large molecular assemblies within one sample that may contain a mixture of complexes in various conformational states. We propose development of dedicated data processing and statistical tools for establishing the number of conformers in the data set, and for studies of conformational modes of the structure, as directly obtained from the EM data.