Experiment data used in DeepHelicon

The data is for the paper titled DeepHelicon: accurate prediction of inter-helical residue contacts in transmembrane protein by residual neural networks.
It contains four sub-folders as follows:
1. Fasta: protein sequences of TRAIN, PREVIOUS, TEST.
2. PDB: protein native structures of TRAIN, PREVIOUS, TEST.
3. Predictions: contact predictions of PREVIOUS and TEST, which are predicted by contact prediction methods mentioned in DeepHelicon paper.
4. 3D modelling: 3D models are generated by CONFOLD2 and guided by residue contacts predicted by DeepHelicon and DeepMetaPSICOV and guided by secondary structures predicted by SCRATCH1D.

本数据集配套论文题为《DeepHelicon：基于残差神经网络（Residual Neural Networks）精准预测跨膜蛋白的螺旋间残基接触》。
本数据集包含如下四个子文件夹：
1. Fasta(Fasta)文件夹：收录训练集（TRAIN）、前驱数据集（PREVIOUS）与测试集（TEST）的蛋白质序列。
2. PDB(PDB)文件夹：收录训练集（TRAIN）、前驱数据集（PREVIOUS）与测试集（TEST）的蛋白质天然结构。
3. 预测结果(Predictions)文件夹：包含前驱数据集（PREVIOUS）与测试集（TEST）的残基接触预测结果，该类结果由《DeepHelicon》论文中提及的接触预测方法生成。
4. 三维建模(3D modelling)文件夹：通过CONFOLD2生成三维模型，建模过程以DeepHelicon与DeepMetaPSICOV预测的残基接触信息、SCRATCH1D预测的二级结构信息作为引导约束。