Building protein structure models for intermediate resolution cryo-electron microscopy maps

项目摘要 冷冻电子显微镜（cryo-EM）是结构生物学中的新兴技术，其能够 确定生物大分子的三维（3D）结构。相比于常规 结构生物学技术，如X射线晶体学和NMR，冷冻EM的主要优点是其 解决大分子组装的能力。此外，最近在冷冻EM方面的技术突破 能够以接近原子级的分辨率确定3D结构。Cryo-EM无疑将成为 在未来十年中，结构生物学的核心重要性。随着冷冻EM的迅速积累 结构化数据，开发能够有效地构建和提取 来自EM图的生物大分子的3D结构。该项目的目标是开发计算 用于建模全局和局部结构以及用于解释嵌入EM中的3D结构的方法 大约4厘米到中等分辨率的地图。最近，我们开发了一种新的从头蛋白质结构， 建模方法MAINMAST，它可以从EM密度图建模蛋白质结构，而无需使用 地图上现有的模板或片段结构。在成功开发MAINMAST的基础上， 进一步扩展MAINMAST的能力，使其更精确地建模和多链建模。在 此外，我们还将开发新的建模方法，为中等分辨率的电磁地图，其中联合收割机， 粗粒度蛋白质结构建模技术，蛋白质结构预测方法，以及低- 分辨率图像处理方法与深度学习，一个国家的最先进的强大的机器学习方法。 拟议项目利用了在结构确定方面所做的巨大努力和取得的进展， 通过开发计算工具，使研究人员能够执行有效和可靠的结构， 分析3D EM密度图。该项目将极大地促进对分子的研究 通过提供3D结构建模的有效手段来研究大分子功能机制。

Project Summary Cryo-electron microscopy (cryo-EM) is an emerging technique in structural biology, which is capable of determining three-dimensional (3D) structures of biological macromolecules. Compared to conventional structural biology techniques, such as X-ray crystallography and NMR, a major advantage of cryo-EM is its ability to solve large macromolecular assemblies. Moreover, recent technical breakthroughs in cryo-EM have enabled determination of 3D structures at nearly atomic-level resolutions. Cryo-EM will undoubtedly become a method of central importance in structural biology in the next decade. With the rapid accumulation of cryo-EM structure data, it has become crucial to develop computational methods that can effectively build and extract 3D structures of biological macromolecules from EM maps. The goal of this project is to develop computational methods for modeling both global and local structures and for interpreting 3D structures embedded in EM maps of around 4 Å to medium-resolution. Recently, we have developed a new de novo protein structure modeling method, MAINMAST, which can model protein structures from an EM density map without using existing template or fragment structures on the map. Based on the successful development of MAINMAST, we further extend the capability of MAINMAST toward more accurate modeling and for multiple-chain modeling. In addition, we will also develop novel modeling methods for medium-resolution EM maps, which combine a coarse-grained protein structure modeling technique, methods in protein structure prediction, and a low- resolution image processing approach with deep learning, a state-of-the-art powerful machine learning method. The proposed project capitalizes on the tremendous efforts and progress made in structural determination with cryo-EM by developing computational tools that allow researchers to perform efficient and reliable structure analyses for 3D EM density maps. The project will greatly facilitate investigation into the molecular mechanisms of macromolecule function by providing an efficient means of 3D structure modeling.