面向大尺度生物图像的高精度三维分子结构重构技术研究

Three-dimensional (3D) reconstruction of Biological Microscopy imaging has been a leading technique in analyzing 3D structures of biomacromolecule. However, current technique for 3D-reconstruction of microscopic imaging has the following deficiencies: Firstly, it is difficult to resolve the problems of missing wedge,highly noisy data, and image distortion, so that high-quality 3D reconstruction cannot be obtained for large scale biological images.Secondly, The procees is very time-consuming and its efficiency is not satisfactory. Facing the above challenges, the following issues are investigated in the project: (1) Alignment Techniques: Different from existing alignment techniques which cannot effectively utilize the information in large scale biological images in their alignment process for 3D reconstruction, new techniques that can effectively utilize some key biological information for improving the reconstruction resolution of 3D reconstruction will be investigated. (2) Advanced projection and reconstruction models, and algorithms: New curviliear projection models that can improve the quality of 3D reconstruction will be investigated; in addition, iterative reconstruction algorithms that can resolve the missing wedge problem and obtain high resolution of 3D reconstruction for biological microscopic images will be investigated. (3) Parallel processing techniques for biological microscopic images 3D reconstruction: the computational structure of the application will be investigated, corresponding schemes for data partition and distribution will be developed, and job and resource distribution and scheduling algorithms will be investigated. High quality 3D reconstruction techniques will be provided for large scale biological images so that atom-level biological molecular 3D structure can be revealed.

生物显微图像三维重构是研究生物大分子三维结构的前沿技术。但是目前显微影像三维重构技术和软件存在诸如难以处理大规模大尺度生物图像中的信息缺失、高噪声、扭曲变形等问题，难以构造高精度的三维重构结果，以及处理速度和效率不能满足科学家的需求等重要缺陷。本项目研究面向大尺度生物图像的高精度分子结构三维重构的关键技术，包括：1，面向大规模生物图像的对位算法：与现存对位算法不同， 本项目研究能很好地识别和利用真实生物样本自身图像的有关信息而增强对位精度的理论与技术；2，面向大尺度生物图像的高精度重构模型与算法：研究能有效解决扭曲变形、信息缺失等问题的曲线投影模型、迭代重构算法；3，针对该应用特征的并行处理技术，包括针对计算结构特征的数据分布、资源分配与调度技术。本项研究将为生物信息处理领域提供高精度的三维重构技术，为揭示原子级的生物分子结构提供理论和技术。

电子断层三维重构技术是近几年迅速发展起来的研究生物大分子纳米分辨率三维结构的前沿技术，是联接细胞生物学和结构分子生物学研究的重要手段。目前电子断层三维重构技术存在如下缺陷：1. 缺乏有效的处理大规模生物图像的对位算法；2. 缺乏高精度的大规模大尺度生物图像的三维重构技术；3. 缺乏针对大规模生物图像的三维重构高效并行处理技术；4. 缺乏面向大规模大尺度生物图像的三维重构软件系统。.本项目在执行期间，从大尺度生物图像精确对位技术、大尺度电子断层重构算法、大规模电子断层重构并行计算和面向大规模大尺度生物显微图像的电子断层重构软件四个方面取得了进展：实现了对于大尺度大规模生物电镜图像的全自动对位算法Markerauto；提出多种能够弥补电子断层三维重构中信息缺失的高精度重构算法FASART、BSIRT、FIRT、ICON以及多轴重构算法AuTom-dualx；利用动态存储机制和任务调度策略实现了在多种并行平台上的生物图像电子断层重构的高效并行处理技术；开发了一套完整的面向大规模大尺度生物显微影像的全自动电子断层重构软件系统。项目实施过程中，主要参与人员至今已经在Cell Research（中科院JCR一区）、Bioinformatics、Journal of Structural Biology、BMC Bioinformatics、IEEE Transactions on Nanobioscience、IEEE/ACM Transactions on Computational Biology and Bioinformatics等国际顶级期刊和会议上发表论文27篇，其中SCI期刊论文20篇，EI会议论文7篇，获得授权的专利两项。