DeepSLICEM: Clustering CryoEM particles using deep image and similarity graph representations: Experiment Data

Details of experiments are given in the paper, titled 'DeepSLICEM: Clustering CryoEM particles using deep image and similarity graph representations'. Supporting code is available on GitHub at: https://github.com/marcottelab/2D_projection_clustering/ Details of the data are given below: Folder CryoEM_data contains 6 datasets - one experimental, and 5 synthetic with 2D projection images of CryoEM particles, common-lines based similarity graph of the images, and ground truth clusters of the images corresponding to the same particles. Folder DeepSLICEM_graph_node_embeddings contains graph node embeddings of the common-lines based similarity graph of the 2D projection images for each of the 6 datasets. Folder DeepSLICEM_clustering_results contains algorithm predicted clusters of 2D projection images belonging to the same particles in each of the 6 datasets, along with the embeddings used for clustering, such as combined image and graph node embeddings. Folder DeepSLICEM_siamese_neural_network_models contains models trained to obtain fine-tuned embeddings of 2D projection images based on similarity w.r.t the same particle's projections.相关实验细节已发表于题为《DeepSLICEM：基于深度图像与相似性图表征的冷冻电镜（CryoEM）颗粒聚类》的论文中。 配套代码已上传至GitHub仓库：https://github.com/marcottelab/2D_projection_clustering/ 数据集详情如下： CryoEM_data 文件夹包含6组数据集——1组实验数据集与5组合成数据集，内含冷冻电镜颗粒的二维投影图像、基于公共线的图像相似性图，以及对应同一颗粒的图像真实聚类簇。 DeepSLICEM_graph_node_embeddings 文件夹收录了6组数据集各自对应的二维投影图像的基于公共线的相似性图的图节点嵌入结果。 DeepSLICEM_clustering_results 文件夹包含6组数据集各自的算法预测聚类结果（即归属同一颗粒的二维投影图像聚类簇），以及聚类所用的嵌入特征，如图像与图节点的融合嵌入。 DeepSLICEM_siamese_neural_network_models 文件夹收录了经训练的孪生神经网络模型，该模型用于基于与同一颗粒投影的相似性生成二维投影图像的微调嵌入表征。