Loading deepchem/feat/atomic_coordinates.py +2 −0 Original line number Diff line number Diff line Loading @@ -227,6 +227,8 @@ class AtomicConvFeaturizer(ComplexFeaturizer): # TODO(rbharath): extend to more fragments if len(frag_num_atoms) != 2: raise ValueError("Currently only supports two fragments") self.frag1_num_atoms = frag_num_atoms[0] self.frag2_num_atoms = frag_num_atoms[0] self.complex_num_atoms = sum(frag_num_atoms) self.max_num_neighbors = max_num_neighbors self.neighbor_cutoff = neighbor_cutoff Loading deepchem/molnet/load_function/pdbbind_datasets.py +391 −4 Original line number Diff line number Diff line Loading @@ -147,6 +147,391 @@ def load_pdbbind_grid(split="random", return tasks, (train, valid, test), transformers def download_pdbbind(data_dir=None, subset="refined", version="v2015", interactions="protein-ligand"): """Downloads PDBBind data to a local directory. This utility function will download the raw PDBBind dataset into a local directory. If you don't specify `data_dir`, it will default to `deepchem.utils.get_data_dir()`. You can download either the 2015 or 2019 versions of PDBBind data with this function by specifying the subset in question. Parameters ---------- data_dir: Str, optional Specifies the directory storing the raw dataset. subset: str, optional This is one of "refined", or "other". These are subsets of the protein-ligand interactions (if `interactions != "protein-ligand"` this field is ignored). If `version=="v2015"`, then this field is ignored since everything is stored in one big gzip together. If `version=="v2019"`, then "refined" and "other" are separate datasets which must be downloaded separately. version: str, optional Either "v2015" or "v2019". Only "v2015" is supported for now. interactions: str, optional Must be one of "protein-ligand", "protein-protein", "protein-nucleic-acid", "nucleic-acid-ligand". If `version=='v2015'`, only "protein-ligand" is supported. """ if version == "v2015" and interactions != "protein-ligand": raise ValueError("Only protein-ligand interactions supported for v2015") if data_dir == None: data_dir = deepchem.utils.get_data_dir() if version == "v2015": data_folder = os.path.join(data_dir, "pdbbind", "v2015") dataset_file = os.path.join(data_dir, "pdbbind_v2015.tar.gz") if not os.path.exists(dataset_file): logger.warning("About to download PDBBind full dataset. Large file, 2GB") deepchem.utils.download_url( 'http://deepchem.io.s3-website-us-west-1.amazonaws.com/datasets/' + "pdbbind_v2015.tar.gz", dest_dir=data_dir) if os.path.exists(data_folder): logger.info("PDBBind v2015 full dataset already exists.") else: logger.info("Untarring PDBBind v2015 full dataset...") deepchem.utils.untargz_file( dataset_file, dest_dir=os.path.join(data_dir, "pdbbind")) elif version == "v2019": data_folder = os.path.join(data_dir, "pdbbind", "v2019") # Download the index file since we'll always need it index_file = os.path.join(data_dir, "PDBbind_2019_plain_text_index.tar.gz") index_folder = os.path.join(data_folder, "plain-text-index") if not os.path.exists(index_file): logger.info("About to download PDBBind 2019 index file.") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/PDBbind_2019_plain_text_index.tar.gz') if not os.path.exists(index_folder): logger.info("Untarring 2019 index dataset...") deepchem.utils.untargz_file( index_file, dest_dir=data_folder) if interactions == "protein-protein": pp_file = os.path.join(data_dir, "pdbbind_v2019_PP.tar.gz") if not os.path.exists(pp_file): logger.warning("About to download PDBBind 2019 protein-protein interactions. Large file of 688 MB") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/pdbbind_v2019_PP.tar.gz') pp_folder = os.path.join(data_folder, "PP") if not os.path.exists(pp_folder): logger.info("Untarring 2019 protein-protein dataset...") deepchem.utils.untargz_file( pp_file, dest_dir=data_folder) elif interactions == "protein-nucleic-acid": pn_file = os.path.join(data_dir, "pdbbind_v2019_PN.tar.gz") if not os.path.exists(pn_file): logger.warning("About to download PDBBind 2019 protein-nucleic-acid interactions. Large file of 229 MB") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/pdbbind_v2019_PN.tar.gz') pn_folder = os.path.join(data_folder, "PN") if not os.path.exists(pn_folder): logger.info("Untarring 2019 protein-nucleic-acid dataset...") deepchem.utils.untargz_file( pn_file, dest_dir=data_folder) elif interactions == "nucleic-acid-ligand": nl_file = os.path.join(data_dir, "pdbbind_v2019_NL.tar.gz") if not os.path.exists(nl_file): logger.warning("About to download PDBBind 2019 nucleic-acid-ligand interactions. File of 17 MB") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/pdbbind_v2019_NL.tar.gz') nl_folder = os.path.join(data_folder, "NL") if not os.path.exists(nl_folder): logger.info("Untarring 2019 nucleic-acid-ligand dataset...") deepchem.utils.untargz_file( nl_file, dest_dir=data_folder) elif interactions == "protein-ligand" and subset=="refined": pl_refined_file = os.path.join(data_dir, "pdbbind_v2019_refined.tar.gz") if not os.path.exists(pl_refined_file): logger.warning("About to download PDBBind 2019 protein-ligand refined interactions. Large File of 622 MB") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/pdbbind_v2019_refined.tar.gz') pl_refined_folder = os.path.join(data_folder, "refined-set") if not os.path.exists(pl_refined_folder): logger.info("Untarring 2019 protein-ligand refined dataset...") deepchem.utils.untargz_file( pl_refined_file, dest_dir=data_folder) elif interactions == "protein-ligand" and subset=="other": pl_other_file = os.path.join(data_dir, "pdbbind_v2019_other_PL.tar.gz") if not os.path.exists(pl_other_file): logger.warning("About to download PDBBind 2019 protein-ligand other interactions. Large File of 1.6 GB") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/pdbbind_v2019_other_PL.tar.gz') pl_other_folder = os.path.join(data_folder, "v2019-other-PL") if not os.path.exists(pl_other_folder): logger.info("Untarring 2019 protein-ligand other dataset...") deepchem.utils.untargz_file( pl_other_file, dest_dir=data_folder) def get_pdbbind_molecular_complex_files(data_dir=None, subset="refined", version="v2015", interactions="protein-ligand", load_binding_pocket=False): """Get a list of the structure files for each dataset. A macromolecular complex is typically specified by one or more molecular structure files (PDB/sdf/mol2). This function returns a list of the structure files associated with various PDBBind datasets. This is needed for downstream processing by ComplexFeaturizers which work from these structure files to produce vectorial or tensorial representations of the data. Parameters ---------- data_dir: Str, optional Specifies the directory storing the raw dataset. subset: str, optional This is one of "core", "refined", "general". These are subsets of the protein-ligand interactions (if `interactions != "protein-ligand"` this field is ignored). Note that v2019 doesn't have an included "core" set. version: str, optional Either "v2015" or "v2019". Only "v2015" is supported for now. interactions: str, optional Must be one of "protein-ligand", "protein-protein", "protein-nucleic-acid", "nucleic-acid-ligand". If `version=='v2015'`, only "protein-ligand" is supported. load_binding_pocket: Bool, optional Load binding pocket or full protein. Only valid for v2015 protein-ligand dataset. Returns ------- List. If a molecular complex has multiple files they are returned together as tuple. """ if data_dir == None: data_dir = deepchem.utils.get_data_dir() if version == "v2015": data_folder = os.path.join(data_dir, "pdbbind", "v2015") if subset == "core": index_labels_file = os.path.join(data_folder, "INDEX_core_data.2013") elif subset == "refined": index_labels_file = os.path.join(data_folder, "INDEX_refined_data.2015") elif subset == "general": index_labels_file = os.path.join(data_folder, "INDEX_general_PL_data.2015") else: raise ValueError("Other subsets not supported") # Extract locations of data with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] if load_binding_pocket: protein_files = [ os.path.join(data_folder, pdb, "%s_pocket.pdb" % pdb) for pdb in pdbs ] else: protein_files = [ os.path.join(data_folder, pdb, "%s_protein.pdb" % pdb) for pdb in pdbs ] ligand_files = [ os.path.join(data_folder, pdb, "%s_ligand.sdf" % pdb) for pdb in pdbs ] return list(zip(protein_files, ligand_files)) elif version == "v2019": data_folder = os.path.join(data_dir, "pdbbind", "v2019") if interactions == "protein-protein": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PP.2019") with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] pp_folder = os.path.join(data_folder, "PP") protein_files = [ os.path.join(pp_folder, "%s.ent.pdb" % pdb) for pdb in pdbs ] return protein_files elif interactions == "protein-nucleic-acid": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PN.2019") with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] pn_folder = os.path.join(data_folder, "PN") protein_files = [ os.path.join(pn_folder, "%s.ent.pdb" % pdb) for pdb in pdbs ] return protein_files elif interactions == "nucleic-acid-ligand": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_NL.2019") with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] nl_folder = os.path.join(data_folder, "NL") complex_files = [ os.path.join(nl_folder, "%s.ent.pdb" % pdb) for pdb in pdbs ] return complex_files elif interactions == "protein-ligand": if subset == "refined": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_refined_data.2019") with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] refined_folder = os.path.join(data_folder, "refined-set") protein_files = [ os.path.join(refined_folder, pdb, "%s_protein.pdb" % pdb) for pdb in pdbs ] ligand_files = [ os.path.join(data_folder, pdb, "%s_ligand.sdf" % pdb) for pdb in pdbs ] return list(zip(protein_files, ligand_files)) elif subset == "general": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PL_data.2019") with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] general_folder = os.path.join(data_folder, "v2019-other-PL") protein_files = [ os.path.join(general_folder, pdb, "%s_protein.pdb" % pdb) for pdb in pdbs ] ligand_files = [ os.path.join(data_folder, pdb, "%s_ligand.sdf" % pdb) for pdb in pdbs ] return list(zip(protein_files, ligand_files)) else: raise ValueError("Other subsets not supported") else: raise ValueError("Only v2015 and v2019 versions are supported.") def get_pdbbind_molecular_complex_labels(data_dir=None, subset="refined", version="v2015", interactions="protein-ligand", load_binding_pocket=False): """Get a list of the labels for each dataset. This function returns a list of the labels associated with various PDBBind datasets. Labels will be returned in the same order as the raw files from `get_pdbbind_molecular_complex_files`. Protein-ligand datasets have units -log Kd/Ki, while the protein-protein, protein-nucleic-acid, nucleic-acid-ligand datasets have units as raw Kd/Ki/IC50 values. These later files are occasionally thresholded, with readings such as 'Ki<1fM' converted to '1fM'. If you'd like to do a more refined conversion, you'll need to refer to the source label file. Parameters ---------- data_dir: Str, optional Specifies the directory storing the raw dataset. subset: str, optional This is one of "core", "refined", "general". These are subsets of the protein-ligand interactions (if `interactions != "protein-ligand"` this field is ignored). Note that v2019 doesn't have an included "core" set. version: str, optional Either "v2015" or "v2019". Only "v2015" is supported for now. interactions: str, optional Must be one of "protein-ligand", "protein-protein", "protein-nucleic-acid", "nucleic-acid-ligand". If `version=='v2015'`, only "protein-ligand" is supported. load_binding_pocket: Bool, optional Load binding pocket or full protein. Only valid for v2015 protein-ligand dataset. Returns ------- List. If a molecular complex has multiple files they are returned together as tuple. """ if data_dir == None: data_dir = deepchem.utils.get_data_dir() if version == "v2015": data_folder = os.path.join(data_dir, "pdbbind", "v2015") if subset == "core": index_labels_file = os.path.join(data_folder, "INDEX_core_data.2013") elif subset == "refined": index_labels_file = os.path.join(data_folder, "INDEX_refined_data.2015") elif subset == "general": index_labels_file = os.path.join(data_folder, "INDEX_general_PL_data.2015") else: raise ValueError("Other subsets not supported") # Extract labels with open(index_labels_file, "r") as g: labels = np.array([ # Lines have format # PDB code, resolution, release year, -logKd/Ki, Kd/Ki, reference, ligand name # The base-10 logarithm, -log kd/pk float(line.split()[3]) for line in g.readlines() if line[0] != "#" ]) return labels elif version == "v2019": data_folder = os.path.join(data_dir, "pdbbind", "v2019") # process interactions that have binding data if interactions in ["protein-protein", "protein-nucleic-acid", "nucleic-acid-ligand"]: if interactions == "protein-protein": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PP.2019") elif interactions == "protein-nucleic-acid": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PN.2019") elif interactions == "nucleic-acid-ligand": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_NL.2019") # Lines have format # PDB code, resolution, release year, binding data, reference, ligand name # Extract labels with open(index_labels_file, "r") as g: raw_labels = [ line.split()[3] for line in g.readlines() if line[0] != "#" ] clean_labels = [] # Here are few types of raw labels we can see in the source # data: Kd=31.8uM, IC50=0.6nM, Kd=22.8pM, Kd~1nM, Kd>500uM, # Ki<0.002nM, Kd=1fM, Kd=3mM. # We have to do two steps of processing. The first is to remove # the separator. Possible values are ["=", "<", "~", ">"]. # For the inequalities, we threshold so "<1fM" becomes "1fM". # This may cause some distortion in learning. separators = ["=", "<", "~", ">"] # The second step we have to do is separate the units unit_conversions = {"mM": 1e-3, "uM":1e-6, "nM":1e-9, "pM":1e-12, "fM":1e-15} number = None for raw in raw_labels: for separator in separators: if separator in raw: pieces = raw.split(separator) # This is something like "0.6nM" number = pieces[-1] break if number is None: raise ValueError("Don't know how to parse %s" % raw) # Shave off units mantissa = float(number[:-2] ) units = unit_conversions[number[-2:]] clean_labels.append(mantissa * units) # Make sure we didn't miss any labels in processing assert len(clean_labels) == len(raw_labels) return clean_labels elif interactions == "protein-ligand": if subset == "refined": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_refined_data.2019") elif subset == "general": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PL_data.2019") else: raise ValueError("Other subsets not supported") # Extract labels with open(index_labels_file, "r") as g: labels = np.array([ # Lines have format # PDB code, resolution, release year, -logKd/Ki, Kd/Ki, reference, ligand name # The base-10 logarithm, -log kd/pk float(line.split()[3]) for line in g.readlines() if line[0] != "#" ]) return labels else: raise ValueError("Only v2015 and v2019 versions are supported.") def load_pdbbind(reload=True, data_dir=None, Loading Loading @@ -176,11 +561,13 @@ def load_pdbbind(reload=True, split_seed: Int, optional Specifies the random seed for splitter. save_dir: Str, optional Specifies the directory to store the featurized and splitted dataset when reload is False. If reload is True, it will load saved dataset inside save_dir. Specifies the directory to store the featurized and splitted dataset when reload is False. If reload is True, it will load saved dataset inside save_dir. save_timestamp: Bool, optional Save featurized and splitted dataset with timestamp or not. Set it as True when running similar or same jobs simultaneously on multiple compute nodes. Save featurized and splitted dataset with timestamp or not. Set it as True when running similar or same jobs simultaneously on multiple compute nodes. """ pdbbind_tasks = ["-logKd/Ki"] Loading examples/pdbbind/atomic_conv_featurization_example.py 0 → 100644 +22 −0 Original line number Diff line number Diff line # This example featurizes a small subset of the PDBBind v2015 protein-ligand core data with the AtomicConvFeaturizer import deepchem as dc from deepchem.molnet.load_function.pdbbind_datasets import get_pdbbind_molecular_complex_files complex_files = get_pdbbind_molecular_complex_files(subset="core", version="v2015", interactions="protein-ligand", load_binding_pocket=False) core_subset = complex_files[:2] ligand_files = [core[1] for core in core_subset] protein_files = [core[0] for core in core_subset] frag1_num_atoms = 70 # for ligand atoms frag2_num_atoms = 24000 # for protein atoms complex_num_atoms = 24070 # in total max_num_neighbors = 4 # Cutoff in angstroms neighbor_cutoff = 4 featurizer = dc.feat.AtomicConvFeaturizer( frag_num_atoms=[frag1_num_atoms,frag2_num_atoms], complex_num_atoms=complex_num_atoms, max_num_neighbors=max_num_neighbors, neighbor_cutoff=neighbor_cutoff) features, failures = featurizer.featurize_complexes( ligand_files, protein_files, parallelize=False) examples/pdbbind/download_pdbbind.py 0 → 100644 +24 −0 Original line number Diff line number Diff line # This example demonstrates how to download the raw PDBBind data import deepchem as dc import logging logging.basicConfig(level=logging.INFO) # Download 2015 protein-ligand dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2015", interactions="protein-ligand") # Download 2019 protein-protein dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2019", interactions="protein-protein") # Download 2019 protein-nucleic-acid dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2019", interactions="protein-nucleic-acid") # Download 2019 nucleic-acid-ligand dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2019", interactions="nucleic-acid-ligand") # Download 2019 protein-ligand refined dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2019", interactions="protein-ligand", subset="refined") # Download 2019 protein-ligand other dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2019", interactions="protein-ligand", subset="other") examples/pdbbind/get_molecular_complex_files.py 0 → 100644 +27 −0 Original line number Diff line number Diff line # This examples shows how to get the list of structure files for PDBBind datasets import deepchem as dc from deepchem.molnet.load_function.pdbbind_datasets import get_pdbbind_molecular_complex_files pdbbind_v2015_core_files = get_pdbbind_molecular_complex_files(subset="core", version="v2015", interactions="protein-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2015 core set: %d" % len(pdbbind_v2015_core_files)) pdbbind_v2015_refined_files = get_pdbbind_molecular_complex_files(subset="refined", version="v2015", interactions="protein-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2015 refined set: %d" % len(pdbbind_v2015_refined_files)) pdbbind_v2015_general_files = get_pdbbind_molecular_complex_files(subset="general", version="v2015", interactions="protein-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2015 general set: %d" % len(pdbbind_v2015_general_files)) pdbbind_v2019_PP_files = get_pdbbind_molecular_complex_files(subset="general", version="v2019", interactions="protein-protein", load_binding_pocket=False) print("Number of files in PDBBind v2019 PP set: %d" % len(pdbbind_v2019_PP_files)) pdbbind_v2019_PN_files = get_pdbbind_molecular_complex_files(subset="general", version="v2019", interactions="protein-nucleic-acid", load_binding_pocket=False) print("Number of files in PDBBind v2019 PN set: %d" % len(pdbbind_v2019_PN_files)) pdbbind_v2019_NL_files = get_pdbbind_molecular_complex_files(subset="general", version="v2019", interactions="nucleic-acid-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2019 NL set: %d" % len(pdbbind_v2019_NL_files)) pdbbind_v2019_refined_files = get_pdbbind_molecular_complex_files(subset="refined", version="v2019", interactions="protein-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2019 refined set: %d" % len(pdbbind_v2019_refined_files)) pdbbind_v2019_general_files = get_pdbbind_molecular_complex_files(subset="general", version="v2019", interactions="protein-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2019 general set: %d" % len(pdbbind_v2019_general_files)) Loading
deepchem/feat/atomic_coordinates.py +2 −0 Original line number Diff line number Diff line Loading @@ -227,6 +227,8 @@ class AtomicConvFeaturizer(ComplexFeaturizer): # TODO(rbharath): extend to more fragments if len(frag_num_atoms) != 2: raise ValueError("Currently only supports two fragments") self.frag1_num_atoms = frag_num_atoms[0] self.frag2_num_atoms = frag_num_atoms[0] self.complex_num_atoms = sum(frag_num_atoms) self.max_num_neighbors = max_num_neighbors self.neighbor_cutoff = neighbor_cutoff Loading
deepchem/molnet/load_function/pdbbind_datasets.py +391 −4 Original line number Diff line number Diff line Loading @@ -147,6 +147,391 @@ def load_pdbbind_grid(split="random", return tasks, (train, valid, test), transformers def download_pdbbind(data_dir=None, subset="refined", version="v2015", interactions="protein-ligand"): """Downloads PDBBind data to a local directory. This utility function will download the raw PDBBind dataset into a local directory. If you don't specify `data_dir`, it will default to `deepchem.utils.get_data_dir()`. You can download either the 2015 or 2019 versions of PDBBind data with this function by specifying the subset in question. Parameters ---------- data_dir: Str, optional Specifies the directory storing the raw dataset. subset: str, optional This is one of "refined", or "other". These are subsets of the protein-ligand interactions (if `interactions != "protein-ligand"` this field is ignored). If `version=="v2015"`, then this field is ignored since everything is stored in one big gzip together. If `version=="v2019"`, then "refined" and "other" are separate datasets which must be downloaded separately. version: str, optional Either "v2015" or "v2019". Only "v2015" is supported for now. interactions: str, optional Must be one of "protein-ligand", "protein-protein", "protein-nucleic-acid", "nucleic-acid-ligand". If `version=='v2015'`, only "protein-ligand" is supported. """ if version == "v2015" and interactions != "protein-ligand": raise ValueError("Only protein-ligand interactions supported for v2015") if data_dir == None: data_dir = deepchem.utils.get_data_dir() if version == "v2015": data_folder = os.path.join(data_dir, "pdbbind", "v2015") dataset_file = os.path.join(data_dir, "pdbbind_v2015.tar.gz") if not os.path.exists(dataset_file): logger.warning("About to download PDBBind full dataset. Large file, 2GB") deepchem.utils.download_url( 'http://deepchem.io.s3-website-us-west-1.amazonaws.com/datasets/' + "pdbbind_v2015.tar.gz", dest_dir=data_dir) if os.path.exists(data_folder): logger.info("PDBBind v2015 full dataset already exists.") else: logger.info("Untarring PDBBind v2015 full dataset...") deepchem.utils.untargz_file( dataset_file, dest_dir=os.path.join(data_dir, "pdbbind")) elif version == "v2019": data_folder = os.path.join(data_dir, "pdbbind", "v2019") # Download the index file since we'll always need it index_file = os.path.join(data_dir, "PDBbind_2019_plain_text_index.tar.gz") index_folder = os.path.join(data_folder, "plain-text-index") if not os.path.exists(index_file): logger.info("About to download PDBBind 2019 index file.") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/PDBbind_2019_plain_text_index.tar.gz') if not os.path.exists(index_folder): logger.info("Untarring 2019 index dataset...") deepchem.utils.untargz_file( index_file, dest_dir=data_folder) if interactions == "protein-protein": pp_file = os.path.join(data_dir, "pdbbind_v2019_PP.tar.gz") if not os.path.exists(pp_file): logger.warning("About to download PDBBind 2019 protein-protein interactions. Large file of 688 MB") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/pdbbind_v2019_PP.tar.gz') pp_folder = os.path.join(data_folder, "PP") if not os.path.exists(pp_folder): logger.info("Untarring 2019 protein-protein dataset...") deepchem.utils.untargz_file( pp_file, dest_dir=data_folder) elif interactions == "protein-nucleic-acid": pn_file = os.path.join(data_dir, "pdbbind_v2019_PN.tar.gz") if not os.path.exists(pn_file): logger.warning("About to download PDBBind 2019 protein-nucleic-acid interactions. Large file of 229 MB") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/pdbbind_v2019_PN.tar.gz') pn_folder = os.path.join(data_folder, "PN") if not os.path.exists(pn_folder): logger.info("Untarring 2019 protein-nucleic-acid dataset...") deepchem.utils.untargz_file( pn_file, dest_dir=data_folder) elif interactions == "nucleic-acid-ligand": nl_file = os.path.join(data_dir, "pdbbind_v2019_NL.tar.gz") if not os.path.exists(nl_file): logger.warning("About to download PDBBind 2019 nucleic-acid-ligand interactions. File of 17 MB") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/pdbbind_v2019_NL.tar.gz') nl_folder = os.path.join(data_folder, "NL") if not os.path.exists(nl_folder): logger.info("Untarring 2019 nucleic-acid-ligand dataset...") deepchem.utils.untargz_file( nl_file, dest_dir=data_folder) elif interactions == "protein-ligand" and subset=="refined": pl_refined_file = os.path.join(data_dir, "pdbbind_v2019_refined.tar.gz") if not os.path.exists(pl_refined_file): logger.warning("About to download PDBBind 2019 protein-ligand refined interactions. Large File of 622 MB") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/pdbbind_v2019_refined.tar.gz') pl_refined_folder = os.path.join(data_folder, "refined-set") if not os.path.exists(pl_refined_folder): logger.info("Untarring 2019 protein-ligand refined dataset...") deepchem.utils.untargz_file( pl_refined_file, dest_dir=data_folder) elif interactions == "protein-ligand" and subset=="other": pl_other_file = os.path.join(data_dir, "pdbbind_v2019_other_PL.tar.gz") if not os.path.exists(pl_other_file): logger.warning("About to download PDBBind 2019 protein-ligand other interactions. Large File of 1.6 GB") deepchem.utils.download_url( 'https://deepchemdata.s3-us-west-1.amazonaws.com/datasets/pdbbindv2019/pdbbind_v2019_other_PL.tar.gz') pl_other_folder = os.path.join(data_folder, "v2019-other-PL") if not os.path.exists(pl_other_folder): logger.info("Untarring 2019 protein-ligand other dataset...") deepchem.utils.untargz_file( pl_other_file, dest_dir=data_folder) def get_pdbbind_molecular_complex_files(data_dir=None, subset="refined", version="v2015", interactions="protein-ligand", load_binding_pocket=False): """Get a list of the structure files for each dataset. A macromolecular complex is typically specified by one or more molecular structure files (PDB/sdf/mol2). This function returns a list of the structure files associated with various PDBBind datasets. This is needed for downstream processing by ComplexFeaturizers which work from these structure files to produce vectorial or tensorial representations of the data. Parameters ---------- data_dir: Str, optional Specifies the directory storing the raw dataset. subset: str, optional This is one of "core", "refined", "general". These are subsets of the protein-ligand interactions (if `interactions != "protein-ligand"` this field is ignored). Note that v2019 doesn't have an included "core" set. version: str, optional Either "v2015" or "v2019". Only "v2015" is supported for now. interactions: str, optional Must be one of "protein-ligand", "protein-protein", "protein-nucleic-acid", "nucleic-acid-ligand". If `version=='v2015'`, only "protein-ligand" is supported. load_binding_pocket: Bool, optional Load binding pocket or full protein. Only valid for v2015 protein-ligand dataset. Returns ------- List. If a molecular complex has multiple files they are returned together as tuple. """ if data_dir == None: data_dir = deepchem.utils.get_data_dir() if version == "v2015": data_folder = os.path.join(data_dir, "pdbbind", "v2015") if subset == "core": index_labels_file = os.path.join(data_folder, "INDEX_core_data.2013") elif subset == "refined": index_labels_file = os.path.join(data_folder, "INDEX_refined_data.2015") elif subset == "general": index_labels_file = os.path.join(data_folder, "INDEX_general_PL_data.2015") else: raise ValueError("Other subsets not supported") # Extract locations of data with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] if load_binding_pocket: protein_files = [ os.path.join(data_folder, pdb, "%s_pocket.pdb" % pdb) for pdb in pdbs ] else: protein_files = [ os.path.join(data_folder, pdb, "%s_protein.pdb" % pdb) for pdb in pdbs ] ligand_files = [ os.path.join(data_folder, pdb, "%s_ligand.sdf" % pdb) for pdb in pdbs ] return list(zip(protein_files, ligand_files)) elif version == "v2019": data_folder = os.path.join(data_dir, "pdbbind", "v2019") if interactions == "protein-protein": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PP.2019") with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] pp_folder = os.path.join(data_folder, "PP") protein_files = [ os.path.join(pp_folder, "%s.ent.pdb" % pdb) for pdb in pdbs ] return protein_files elif interactions == "protein-nucleic-acid": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PN.2019") with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] pn_folder = os.path.join(data_folder, "PN") protein_files = [ os.path.join(pn_folder, "%s.ent.pdb" % pdb) for pdb in pdbs ] return protein_files elif interactions == "nucleic-acid-ligand": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_NL.2019") with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] nl_folder = os.path.join(data_folder, "NL") complex_files = [ os.path.join(nl_folder, "%s.ent.pdb" % pdb) for pdb in pdbs ] return complex_files elif interactions == "protein-ligand": if subset == "refined": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_refined_data.2019") with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] refined_folder = os.path.join(data_folder, "refined-set") protein_files = [ os.path.join(refined_folder, pdb, "%s_protein.pdb" % pdb) for pdb in pdbs ] ligand_files = [ os.path.join(data_folder, pdb, "%s_ligand.sdf" % pdb) for pdb in pdbs ] return list(zip(protein_files, ligand_files)) elif subset == "general": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PL_data.2019") with open(index_labels_file, "r") as g: pdbs = [line[:4] for line in g.readlines() if line[0] != "#"] general_folder = os.path.join(data_folder, "v2019-other-PL") protein_files = [ os.path.join(general_folder, pdb, "%s_protein.pdb" % pdb) for pdb in pdbs ] ligand_files = [ os.path.join(data_folder, pdb, "%s_ligand.sdf" % pdb) for pdb in pdbs ] return list(zip(protein_files, ligand_files)) else: raise ValueError("Other subsets not supported") else: raise ValueError("Only v2015 and v2019 versions are supported.") def get_pdbbind_molecular_complex_labels(data_dir=None, subset="refined", version="v2015", interactions="protein-ligand", load_binding_pocket=False): """Get a list of the labels for each dataset. This function returns a list of the labels associated with various PDBBind datasets. Labels will be returned in the same order as the raw files from `get_pdbbind_molecular_complex_files`. Protein-ligand datasets have units -log Kd/Ki, while the protein-protein, protein-nucleic-acid, nucleic-acid-ligand datasets have units as raw Kd/Ki/IC50 values. These later files are occasionally thresholded, with readings such as 'Ki<1fM' converted to '1fM'. If you'd like to do a more refined conversion, you'll need to refer to the source label file. Parameters ---------- data_dir: Str, optional Specifies the directory storing the raw dataset. subset: str, optional This is one of "core", "refined", "general". These are subsets of the protein-ligand interactions (if `interactions != "protein-ligand"` this field is ignored). Note that v2019 doesn't have an included "core" set. version: str, optional Either "v2015" or "v2019". Only "v2015" is supported for now. interactions: str, optional Must be one of "protein-ligand", "protein-protein", "protein-nucleic-acid", "nucleic-acid-ligand". If `version=='v2015'`, only "protein-ligand" is supported. load_binding_pocket: Bool, optional Load binding pocket or full protein. Only valid for v2015 protein-ligand dataset. Returns ------- List. If a molecular complex has multiple files they are returned together as tuple. """ if data_dir == None: data_dir = deepchem.utils.get_data_dir() if version == "v2015": data_folder = os.path.join(data_dir, "pdbbind", "v2015") if subset == "core": index_labels_file = os.path.join(data_folder, "INDEX_core_data.2013") elif subset == "refined": index_labels_file = os.path.join(data_folder, "INDEX_refined_data.2015") elif subset == "general": index_labels_file = os.path.join(data_folder, "INDEX_general_PL_data.2015") else: raise ValueError("Other subsets not supported") # Extract labels with open(index_labels_file, "r") as g: labels = np.array([ # Lines have format # PDB code, resolution, release year, -logKd/Ki, Kd/Ki, reference, ligand name # The base-10 logarithm, -log kd/pk float(line.split()[3]) for line in g.readlines() if line[0] != "#" ]) return labels elif version == "v2019": data_folder = os.path.join(data_dir, "pdbbind", "v2019") # process interactions that have binding data if interactions in ["protein-protein", "protein-nucleic-acid", "nucleic-acid-ligand"]: if interactions == "protein-protein": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PP.2019") elif interactions == "protein-nucleic-acid": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PN.2019") elif interactions == "nucleic-acid-ligand": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_NL.2019") # Lines have format # PDB code, resolution, release year, binding data, reference, ligand name # Extract labels with open(index_labels_file, "r") as g: raw_labels = [ line.split()[3] for line in g.readlines() if line[0] != "#" ] clean_labels = [] # Here are few types of raw labels we can see in the source # data: Kd=31.8uM, IC50=0.6nM, Kd=22.8pM, Kd~1nM, Kd>500uM, # Ki<0.002nM, Kd=1fM, Kd=3mM. # We have to do two steps of processing. The first is to remove # the separator. Possible values are ["=", "<", "~", ">"]. # For the inequalities, we threshold so "<1fM" becomes "1fM". # This may cause some distortion in learning. separators = ["=", "<", "~", ">"] # The second step we have to do is separate the units unit_conversions = {"mM": 1e-3, "uM":1e-6, "nM":1e-9, "pM":1e-12, "fM":1e-15} number = None for raw in raw_labels: for separator in separators: if separator in raw: pieces = raw.split(separator) # This is something like "0.6nM" number = pieces[-1] break if number is None: raise ValueError("Don't know how to parse %s" % raw) # Shave off units mantissa = float(number[:-2] ) units = unit_conversions[number[-2:]] clean_labels.append(mantissa * units) # Make sure we didn't miss any labels in processing assert len(clean_labels) == len(raw_labels) return clean_labels elif interactions == "protein-ligand": if subset == "refined": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_refined_data.2019") elif subset == "general": index_labels_file = os.path.join(data_folder, "plain-text-index", "index", "INDEX_general_PL_data.2019") else: raise ValueError("Other subsets not supported") # Extract labels with open(index_labels_file, "r") as g: labels = np.array([ # Lines have format # PDB code, resolution, release year, -logKd/Ki, Kd/Ki, reference, ligand name # The base-10 logarithm, -log kd/pk float(line.split()[3]) for line in g.readlines() if line[0] != "#" ]) return labels else: raise ValueError("Only v2015 and v2019 versions are supported.") def load_pdbbind(reload=True, data_dir=None, Loading Loading @@ -176,11 +561,13 @@ def load_pdbbind(reload=True, split_seed: Int, optional Specifies the random seed for splitter. save_dir: Str, optional Specifies the directory to store the featurized and splitted dataset when reload is False. If reload is True, it will load saved dataset inside save_dir. Specifies the directory to store the featurized and splitted dataset when reload is False. If reload is True, it will load saved dataset inside save_dir. save_timestamp: Bool, optional Save featurized and splitted dataset with timestamp or not. Set it as True when running similar or same jobs simultaneously on multiple compute nodes. Save featurized and splitted dataset with timestamp or not. Set it as True when running similar or same jobs simultaneously on multiple compute nodes. """ pdbbind_tasks = ["-logKd/Ki"] Loading
examples/pdbbind/atomic_conv_featurization_example.py 0 → 100644 +22 −0 Original line number Diff line number Diff line # This example featurizes a small subset of the PDBBind v2015 protein-ligand core data with the AtomicConvFeaturizer import deepchem as dc from deepchem.molnet.load_function.pdbbind_datasets import get_pdbbind_molecular_complex_files complex_files = get_pdbbind_molecular_complex_files(subset="core", version="v2015", interactions="protein-ligand", load_binding_pocket=False) core_subset = complex_files[:2] ligand_files = [core[1] for core in core_subset] protein_files = [core[0] for core in core_subset] frag1_num_atoms = 70 # for ligand atoms frag2_num_atoms = 24000 # for protein atoms complex_num_atoms = 24070 # in total max_num_neighbors = 4 # Cutoff in angstroms neighbor_cutoff = 4 featurizer = dc.feat.AtomicConvFeaturizer( frag_num_atoms=[frag1_num_atoms,frag2_num_atoms], complex_num_atoms=complex_num_atoms, max_num_neighbors=max_num_neighbors, neighbor_cutoff=neighbor_cutoff) features, failures = featurizer.featurize_complexes( ligand_files, protein_files, parallelize=False)
examples/pdbbind/download_pdbbind.py 0 → 100644 +24 −0 Original line number Diff line number Diff line # This example demonstrates how to download the raw PDBBind data import deepchem as dc import logging logging.basicConfig(level=logging.INFO) # Download 2015 protein-ligand dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2015", interactions="protein-ligand") # Download 2019 protein-protein dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2019", interactions="protein-protein") # Download 2019 protein-nucleic-acid dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2019", interactions="protein-nucleic-acid") # Download 2019 nucleic-acid-ligand dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2019", interactions="nucleic-acid-ligand") # Download 2019 protein-ligand refined dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2019", interactions="protein-ligand", subset="refined") # Download 2019 protein-ligand other dataset dc.molnet.load_function.pdbbind_datasets.download_pdbbind(version="v2019", interactions="protein-ligand", subset="other")
examples/pdbbind/get_molecular_complex_files.py 0 → 100644 +27 −0 Original line number Diff line number Diff line # This examples shows how to get the list of structure files for PDBBind datasets import deepchem as dc from deepchem.molnet.load_function.pdbbind_datasets import get_pdbbind_molecular_complex_files pdbbind_v2015_core_files = get_pdbbind_molecular_complex_files(subset="core", version="v2015", interactions="protein-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2015 core set: %d" % len(pdbbind_v2015_core_files)) pdbbind_v2015_refined_files = get_pdbbind_molecular_complex_files(subset="refined", version="v2015", interactions="protein-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2015 refined set: %d" % len(pdbbind_v2015_refined_files)) pdbbind_v2015_general_files = get_pdbbind_molecular_complex_files(subset="general", version="v2015", interactions="protein-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2015 general set: %d" % len(pdbbind_v2015_general_files)) pdbbind_v2019_PP_files = get_pdbbind_molecular_complex_files(subset="general", version="v2019", interactions="protein-protein", load_binding_pocket=False) print("Number of files in PDBBind v2019 PP set: %d" % len(pdbbind_v2019_PP_files)) pdbbind_v2019_PN_files = get_pdbbind_molecular_complex_files(subset="general", version="v2019", interactions="protein-nucleic-acid", load_binding_pocket=False) print("Number of files in PDBBind v2019 PN set: %d" % len(pdbbind_v2019_PN_files)) pdbbind_v2019_NL_files = get_pdbbind_molecular_complex_files(subset="general", version="v2019", interactions="nucleic-acid-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2019 NL set: %d" % len(pdbbind_v2019_NL_files)) pdbbind_v2019_refined_files = get_pdbbind_molecular_complex_files(subset="refined", version="v2019", interactions="protein-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2019 refined set: %d" % len(pdbbind_v2019_refined_files)) pdbbind_v2019_general_files = get_pdbbind_molecular_complex_files(subset="general", version="v2019", interactions="protein-ligand", load_binding_pocket=False) print("Number of files in PDBBind v2019 general set: %d" % len(pdbbind_v2019_general_files))
