Three-Dimensional Reconstruction from Electron Micrographs of Randomly Oriented Macromolecules

随机取向大分子电子显微图像的三维重建。了解大分子的结构对于了解它们在生物系统中的功能是必不可少的。许多大分子只能用三维电子显微镜来分析。对于单一的不对称粒子，与晶体对象相比，三维分辨率被限制在2-3纳米。这里提出的技术将有助于更充分地利用显微图像中包含的信息，是实现更高分辨率的重要一步，从而更好地了解生物功能。提出了一种由随机取向的单分子电子显微图像进行三维重建的方法。生物粒子对辐射敏感，成像必须在低剂量条件下进行。只有将许多相同粒子的信息结合在一起的技术才能克服辐射敏感性的限制。这种方法中最成功的一种是由P.I.开发的SECHET(单次曝光圆锥重建技术)。这种方法的程序已经分发到世界各地的实验室，并在过去七年中用于确定15个以上的结构。然而，SECRET要求重建中使用的所有粒子在样本平面内具有相同的方向。偏离这一首选方向限制了三维分辨率。本文提出了一种根据粒子投影确定粒子取向的新方法。这项技术可用于改进用于保密的投影的取向，并用于确定具有未知取向的粒子的投影角。这项技术的成功开发将大大减少实验工作量，并提高可实现的三维分辨率。Radon变换理论是三维重建算法的基础。大多数研究都集中在设计高效的反演算法上，只有少数应用程序将其作为图像处理工具的潜力。在该方案中，Radon变换被用作三维投影的同时平移和旋转对齐的工具。一旦知道投影的方向，就可以计算三维重建。对准/重建方法将与吉布斯采样技术相结合，以克服噪声限制和图像分类。已经编写了评估拟议技术可行性的程序，并公布了第一批结果。这项新技术将应用于两个对象，肌浆网的钙释放通道和大肠杆菌的50S核糖体亚基。之所以选择这两个大分子，是因为它们展示了有利于测试程序的特征，而且更高分辨率的结果将有助于更好地了解它们的功能。钙释放通道的四重对称性可以用来提高信噪比，而50S核糖体亚基是一个球状结构，几乎没有突起，是测试该方法适用范围的理想方法。

Three-dimensional Reconstruction from Electron Micrographs of Randomly Oriented Macromolecules. Knowledge of the structure of macromolecules is essential to an understanding of their functions in biological systems. Many large macromolecules can only be analyzed by threedimensional electron microscopy. For single asymmetrical particles, in contrast to crystalline objects, resolutions achieved in three dimensions have been limited to 2-3 nm. The techniques proposed here will contribute to a fuller usage of information contained in a micrograph and are an important step toward achieving higher resolution and thus a better understanding of biological function. A method is proposed for three-dimensional reconstruction from electron micrographs of randomly oriented single macromolecules. Biological particles are radiation sensitive, and imaging must be done under low-dose conditions. Only techniques that combine information of many identical particles can overcome the limits imposed by radiation sensitivity. One of the most successful methods of this kind was SECReT (Single Exposure Conical Reconstruction Technique), developed by the P.I.. Programs for this method have been distributed to laboratories world wide and have been used for the determination of more than 15 structures over the last seven years. However, SECReT requires that all particles used in a reconstruction have the same orientation within the plane of the specimen. Deviations from this preferred orientation limit the resolution in three dimensions. Proposed here is a new method for determining the orientation of particles from their projections. This technique can be used for refining the orientation of projections used for SECReT and for the determination of projection angles of particles with unknown orientations. Successful development of this technique will lead to substantial reduction in experimental effort and increase the resolution achievable in three dimensions. The theory of Radon t ransforms is the basis for three-dimensional reconstruction algorithms. Most research has focused on the design of efficient inversion algorithms and only a few applications have made use of their potential as image processing tools. In this proposal Radon transforms are used as tools for simultaneous translational and rotational alignment of projections in three dimensions. Once the orientations of the projections are known a threedimensional reconstruction can be calculated. The alignment/reconstruction method will be combined with a Gibbs sampling technique for overcoming noise limitations and for image classification. Programs to assess the feasibility of the proposed technique have been written and first results have been published. The new technique will be applied to two objects, the calcium release channel from sarcoplasmic reticulum and the 50S ribosomal subunit from Escherichia coli. The two macromolecules were chosen because they exhibit features that are advantageous for testing the procedures, and because results at higher resolution will contribute to a better understanding of their function. The four-fold symmetry of the calcium release channel can be used to increase the signal to noise ratio, whereas the 50S ribosomal subunit is a globular structure with few protuberances and ideal for testing the limits of the applicability of this method.