Data for Exploring Conformational Landscape of Cryo-EM Using Energy-Aware Pathfinding Algorithm

There are four experiments. The main.ipynb notebook in each experiment directory details the pathfinding process and evaluates different algorithms. Hsp90 (PDB ID: 2CG9): A synthetic dataset with two degrees of conformational changes. The .mrcs and .star files (convert into .pkl file first use cryodrgn functions) serve as training inputs for the cryoDRGN model. NLRP3 (PDB ID: 6NPY): A synthetic dataset with three degrees of conformational changes. The .mrcs and .star files (convert into .pkl file first use cryodrgn functions) serve as training inputs for the cryoDRGN model. EMPIAR-10076 in original latent space. EMPIAR-10076 with Analysis Landscape Pipeline: We also adhere to the analyze landscape pipeline designed by the cryoDRGN team. The Jupyter notebooks in that directory illustrate our implementation of this pipeline for the EMPIAR-10076 dataset, starting with 1_sketching.ipynb and concluding in 2_learning_mapping.ipynb. In the EMPIAR-10076 experiment, we utilize the training weights provided by the official cryoDRGN team. The complete weights and further information can be found here Additionally, we filter out states unrelated to conformational changes, following the guidance of the labels provided by the cryoDRGN team here.本数据集包含四组实验。每个实验目录下的main.ipynb脚本详细阐述了路径查找流程，并对不同算法进行了评估。热休克蛋白90（Hsp90，PDB ID: 2CG9）：构象变化程度分为两类的合成数据集。其.mrcs与.star文件（需先使用cryoDRGN工具函数转换为.pkl文件）可作为cryoDRGN模型的训练输入数据。NLRP3（PDB ID: 6NPY）：构象变化程度分为三类的合成数据集。其.mrcs与.star文件（需先使用cryoDRGN工具函数转换为.pkl文件）可作为cryoDRGN模型的训练输入数据。EMPIAR-10076 原始潜空间数据集；以及采用分析景观流程（Analysis Landscape Pipeline）处理后的EMPIAR-10076数据集：本研究同时遵循cryoDRGN团队设计的分析景观流程开展实验。该目录下的Jupyter笔记本展示了我们针对EMPIAR-10076数据集实现该流程的完整过程，从1_sketching.ipynb起始，至2_learning_mapping.ipynb结束。在EMPIAR-10076实验中，我们使用了cryoDRGN官方团队提供的训练权重。完整权重及更多相关信息可在此处获取。此外，我们参考cryoDRGN团队提供的标签指引，过滤掉了与构象变化无关的样本状态。