生物分子纤维的冷冻电镜图像分析技术研究

Cryo-EM technique boosts the development of structural biology and is a strong method that can be an alternative way to crystallography. The improvement of hardware and software make this technique can push the structure resolution of the biological molecule to near-atomic resolution. While the helical filaments formed from biological molecules is a special object of single particle analysis, which based on cryo-EM. Filaments, which are different to the isolated molecules in the ice layer, need to a special process before reconstruction. The precision helical parameters strongly needed in order to achieve high-resolution structure. However, the heterogeneity makes the trouble in helical parameters determination. Therefore, this project focuses on this issue and try to solve the problem through the image processing and machine learning (deep learning). It will benefit the reconstruction of high-resolution filament structure.

冷冻电子显微镜技术促进了结构生物学的发展，生物分子结构解析不在局限于晶体学方法。软件和硬件技术的更新迭代使得基于冷冻电子显微镜的单颗粒分析技术能够解析出近原子分辨率的结构。生物分子形成的螺旋体纤维，是单颗粒分析技术中特殊的对象。不同于冰层中的孤立的分子，纤维重构需要测定其精确的螺旋参数，方可到达高分辨率的结构。为此，螺旋参数是纤维结构重构的关键因素。然而，纤维的多态性，以及样品的不均一性给参数测定带来极大的困扰。因此，本项目通过图像处理，机器学习（深度学习）等技术解决重构中的螺旋参数问题。本项目研究将有利于纤维结构重构，以期获得高分辨率的结构。