FEater: A large-scale 3D molecular conformation dataset to measure the geometric invariance of models

This repository contains the dataset present in the work "FEater: A large-scale 3D molecular conformation dataset to measure the geometric invariance of models". FEater_XXX_PDBSRC contains the source PDB fragments each contains over 8 million fragments. FEater_XXX_PDB contains the three HDF files with of the coordinate as point cloud. It also includes the topology of the fragments. FEater_Minisets contains a subset of features used in the training process. For more details about the FEater_Minisets, please check the FEater_paper.pdf for further details.本仓库包含论文《FEater：用于评估模型几何不变性的大规模三维分子构象数据集》所使用的数据集。 FEater_XXX_PDBSRC 包含源PDB（Protein Data Bank）片段，单份数据集内含超800万条片段。 FEater_XXX_PDB 包含三份以点云形式存储坐标的HDF（Hierarchical Data Format）文件，同时涵盖了片段的拓扑结构信息。 FEater_Minisets 包含训练过程中使用的特征子集。若需了解FEater_Minisets的更多细节，请查阅FEater_paper.pdf获取进一步信息。