GAIBind: A Geometric Attention Model with Interaction-Aware Modeling for Accurate Protein–Ligand Binding Pose and Affinity Prediction

The dataset used in the paper "GAIBind: A Geometric Attention Model with Interaction-Aware Modeling for Accurate Protein–Ligand Binding Pose and Affinity Prediction" is provided here. benchmark.zip Contains the docking and affinity benchmark data used for evaluating GAIBind. This package includes the protein pockets, ligand structures, and processed docking inputs required to reproduce the pose prediction and affinity results reported in our paper. gaibind_train.tar.gz: Contains prepared training data for GAIBind. poses/ contains all structures needed for training poses/ligand contains the ligand conformations used during training poses/ligand/rcsb contains the conformations of the reference ligands poses/pocket contains all pockets extracted from the raw PDB structures poses/uff contains ligand conformations minimized using UFF train/ contains the training CSV