CATH 4.2

该数据集包含了按照CATH拓扑分类法归类的蛋白质结构。该数据集被用于评估PiFold算法的性能，并且为了确保公平比较，采用了与GraphTrans和GVP相似的 数据划分方式。数据规模方面，训练集包含了18024个蛋白质，验证集有608个蛋白质，测试集则有1120个蛋白质。该数据集的任务是蛋白质逆折叠。

This dataset contains protein structures classified via the CATH topology classification system. It is utilized to evaluate the performance of the PiFold algorithm, and to ensure fair benchmarking, it employs a data splitting strategy analogous to that used for GraphTrans and GVP. Regarding dataset size, the training set comprises 18,024 proteins, the validation set has 608 proteins, and the test set includes 1,120 proteins. The core task of this dataset is protein inverse folding.