Ligands detected by CheckMyBlob in the Protein Data Bank

A dataset containing numerical descriptions all ligands that CheckMyBlob was capable of detecting automatically on the entire PDB as of May 1st, 2017. It is the "master" data set containing all ligands queried and detected as described in the Kowiel et al. paper "Automatic recognition of ligands in electron density by machine learning methods".

The file is compressed using 7zip to allow for faster downloads. The compressed file weighs around 1.1 GB, whereas the uncompressed CSV will take close to 3.0 GB of disk space. The all_summary.csv file can be used to reproduce the filtered ligand data sets (CMB, TAMC, CL) described in the Kowiel et al. paper. Additionally, the data set can be used as a source to create data sets based on other filtering criteria (e.g. ligand subsets of your choice), or on its own a as source of knowledge about all ligands that CheckMyBlob was capable of detecting automatically on the entire PDB as of May 1st, 2017.

For machine learning applications, please read the code at https://github.com/dabrze/CheckMyBlob, to see examples describing how to preprocess the data. In particular, the following attributes should not be used during model training or testing: "pdb_code", "res_id", "chain_id", "local_res_atom_count", "local_res_atom_non_h_count", "local_res_atom_non_h_occupancy_sum", "local_res_atom_non_h_electron_sum", "local_res_atom_non_h_electron_occupancy_sum", "local_res_atom_C_count", "local_res_atom_N_count", "local_res_atom_O_count", "local_res_atom_S_count", "dict_atom_non_h_count", "dict_atom_non_h_electron_sum", "dict_atom_C_count", "dict_atom_N_count", "dict_atom_O_count", "dict_atom_S_count", "fo_col", "fc_col", "weight_col", "grid_space", "solvent_radius", "solvent_opening_radius", "part_step_FoFc_std_min", "part_step_FoFc_std_max", "part_step_FoFc_std_step", "local_volume", "res_coverage", "blob_coverage", "blob_volume_coverage", "blob_volume_coverage_second", "res_volume_coverage", "res_volume_coverage_second", "skeleton_data", "resolution_max_limit", "part_step_FoFc_std_min", "part_step_FoFc_std_max", "part_step_FoFc_std_step".

The target attribute for classification is: res_name.