DeepDock training dataset

Data set used to train and evaluate DeepDock. The file contains information of ~16K protein ligand complexes extracted from the general set of PDBbind database v2019. They have been preprocessed using MaSSIF and are ready to be used as input using Pytorch geometric. The target is represented by a mesh, each node contains 4 physchem properties. The ligand is represented as a 2D graph where node features represent atom type and edge features represent bond types.