torchdrug.datasets

Knowledge Graph Datasets

FB15k

class FB15k(path, verbose=1)

Subset of Freebase knowledge base for knowledge graph reasoning.

Statistics:

• #Entity: 14,951

• #Relation: 1,345

• #Triplet: 592,213

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

FB15k237

class FB15k237(path, verbose=1)

A filtered version of FB15k dataset without trivial cases.

Statistics:

• #Entity: 14,541

• #Relation: 237

• #Triplet: 310,116

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

WN18

class WN18(path, verbose=1)

WordNet knowledge base.

Statistics:

• #Entity: 40,943

• #Relation: 18

• #Triplet: 151,442

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

WN18RR

class WN18RR(path, verbose=1)

A filtered version of WN18 dataset without trivial cases.

Statistics:

• #Entity: 40,943

• #Relation: 11

• #Triplet: 93,003

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

Hetionet

class Hetionet(path, verbose=1)

Hetionet for knowledge graph reasoning.

Statistics:

• #Entity: 45,158

• #Relation: 24

• #Triplet: 2,025,177

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

Molecule Property Prediction Datasets

BACE

class BACE(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Binary binding results for a set of inhibitors of human \(\beta\)-secretase 1(BACE-1).

Statistics:

• #Molecule: 1,513

• #Classification task: 1

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

BBBP

class BBBP(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Binary labels of blood-brain barrier penetration.

Statistics:

• #Molecule: 2,039

• #Classification task: 1

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

CEP

class CEP(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Photovoltaic efficiency estimated by Havard clean energy project.

Statistics:

• #Molecule: 20,000

• #Regression task: 1

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

ChEMBLFiltered

class ChEMBLFiltered(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Statistics:

• #Molecule: 430,710

• #Regression task: 1,310

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

ClinTox

class ClinTox(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Qualitative data of drugs approved by the FDA and those that have failed clinical trials for toxicity reasons.

Statistics:

• #Molecule: 1,478

• #Classification task: 2

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Delaney

class Delaney(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Log-scale water solubility of molecules.

Statistics:

• #Molecule: 1,128

• #Regression task: 1

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

FreeSolv

class FreeSolv(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Experimental and calculated hydration free energy of small molecules in water.

Statistics:

• #Molecule: 642

• #Regression task: 1

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

HIV

class HIV(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Experimentally measured abilities to inhibit HIV replication.

Statistics:

• #Molecule: 41,127

• #Classification task: 1

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Lipophilicity

class Lipophilicity(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Experimental results of octanol/water distribution coefficient (logD at pH 7.4).

Statistics:

• #Molecule: 4,200

• #Regression task: 1

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

MUV

class MUV(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Subset of PubChem BioAssay by applying a refined nearest neighbor analysis.

Statistics:

• #Molecule: 93,087

• #Classification task: 17

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Malaria

class Malaria(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Half-maximal effective concentration (EC50) against a parasite that causes malaria.

Statistics:

• #Molecule: 10,000

• #Regression task: 1

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

OPV

class OPV(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Quantum mechanical calculations on organic photovoltaic candidate molecules.

Statistics:

• #Molecule: 94,576

• #Regression task: 8

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

QM8

class QM8(path, node_position=False, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Electronic spectra and excited state energy of small molecules.

Statistics:

• #Molecule: 21,786

• #Regression task: 12

Parameters

• path (str) – path to store the dataset

• node_position (bool, optional) – load node position or not. This will add node_position as a node attribute to each sample.

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

QM9

class QM9(path, node_position=False, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Geometric, energetic, electronic and thermodynamic properties of DFT-modeled small molecules.

Statistics:

• #Molecule: 133,885

• #Regression task: 12

Parameters

• path (str) – path to store the dataset

• node_position (bool, optional) – load node position or not. This will add node_position as a node attribute to each sample.

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

SIDER

class SIDER(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Marketed drugs and adverse drug reactions (ADR) dataset, grouped into 27 system organ classes.

Statistics:

• #Molecule: 1,427

• #Classification task: 27

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Tox21

class Tox21(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Qualitative toxicity measurements on 12 biological targets, including nuclear receptors and stress response pathways.

Statistics:

• #Molecule: 7,831

• #Classification task: 12

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

ToxCast

class ToxCast(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Toxicology data based on in vitro high-throughput screening.

Statistics:

• #Molecule: 8,575

• #Classification task: 617

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

ZINC250k

class ZINC250k(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Subset of ZINC compound database for virtual screening.

Statistics:

• #Molecule: 498,910

• #Regression task: 2

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

ZINC2m

class ZINC2m(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

ZINC compound database for virtual screening. This dataset doesn’t contain any label information.

Statistics:

• #Molecule: 2,000,000

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

MOSES

class MOSES(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Subset of ZINC database for molecule generation. This dataset doesn’t contain any label information.

Statistics:

• #Molecule: 1,936,963

Parameters

• path (str) – path for the CSV dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

PCQM4M

class PCQM4M(path, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Quantum chemistry dataset originally curated under the PubChemQC of molecules.

Statistics:

• #Molecule: 3,803,453

• #Regression task: 1

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Protein Property Prediction Datasets

BetaLactamase

class BetaLactamase(path, verbose=1, sequence_field='primary', number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

The activity values of first-order mutants of the TEM-1 beta-lactamase protein.

Statistics:

• #Train: 4,158

• #Valid: 520

• #Test: 520

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• sequence_field (str, optional) – name of the field of protein sequence in lmdb files

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Fluorescence

class Fluorescence(path, verbose=1, sequence_field='primary', number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

The fitness values of a set of green fluorescent protein mutants.

Statistics:

• #Train: 21,446

• #Valid: 5,362

• #Test: 27,217

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• sequence_field (str, optional) – name of the field of protein sequence in lmdb files

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Stability

class Stability(path, verbose=1, sequence_field='primary', number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

The stability values of proteins under natural environment.

Statistics:

• #Train: 53,571

• #Valid: 2,512

• #Test: 12,851

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• sequence_field (str, optional) – name of the field of protein sequence in lmdb files

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Solubility

class Solubility(path, verbose=1, sequence_field='primary', number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Proteins with binary labels indicating their solubility.

Statistics:

• #Train: 62,478

• #Valid: 6,942

• #Test: 1,999

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• sequence_field (str, optional) – name of the field of protein sequence in lmdb files

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

BinaryLocalization

class BinaryLocalization(path, verbose=1, sequence_field='primary', number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Simpler version of the Subcellular Localization with binary labels indicating whether a protein is membrane-bound or soluble.

Statistics:

• #Train: 5,161

• #Valid: 1,727

• #Test: 1,746

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• sequence_field (str, optional) – name of the field of protein sequence in lmdb files

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

SubcellularLocalization

class SubcellularLocalization(path, verbose=1, sequence_field='primary', number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Class labels indicating where a natural protein locates in the cell.

Statistics:

• #Train: 8,945

• #Valid: 2,248

• #Test: 2,768

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• sequence_field (str, optional) – name of the field of protein sequence in lmdb files

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

EnzymeCommission

class EnzymeCommission(path, test_cutoff=0.95, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

A set of proteins with their 3D structures and EC numbers, which describes their catalysis of biochemical reactions.

Statistics (test_cutoff=0.95):

• #Train: 15,011

• #Valid: 1,664

• #Test: 1,840

Parameters

• path (str) – the path to store the dataset

• test_cutoff (float, optional) – the test cutoff used to split the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

GeneOntology

class GeneOntology(path, branch='MF', test_cutoff=0.95, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

A set of proteins with their 3D structures and GO terms. These terms classify proteins into hierarchically related functional classes organized into three ontologies: molecular function (MF), biological process (BP) and cellular component (CC).

Statistics (test_cutoff=0.95):

• #Train: 27,496

• #Valid: 3,053

• #Test: 2,991

Parameters

• path (str) – the path to store the dataset

• branch (str, optional) – the GO branch

• test_cutoff (float, optional) – the test cutoff used to split the dataset

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

AlphaFoldDB

class AlphaFoldDB(path, species_id=0, split_id=0, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

3D protein structures predicted by AlphaFold. This dataset covers proteomes of 48 organisms, as well as the majority of Swiss-Prot.

Statistics:

See https://alphafold.ebi.ac.uk/download

Parameters

• path (str) – path to store the dataset

• species_id (int, optional) – the id of species to be loaded. The species are numbered by the order appeared on https://alphafold.ebi.ac.uk/download (0-20 for model organism proteomes, 21 for Swiss-Prot)

• split_id (int, optional) – the id of split to be loaded. To avoid large memory consumption for one dataset, we have cut each species into several splits, each of which contains at most 22000 proteins.

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Protein Structure Prediction Datasets

Fold

class Fold(path, verbose=1, sequence_field='primary', number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Fold labels for a set of proteins determined by the global structural topology.

Statistics:

• #Train: 12,312

• #Valid: 736

• #Test: 718

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• sequence_field (str, optional) – name of the field of protein sequence in lmdb files

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

SecondaryStructure

class SecondaryStructure(path, verbose=1, sequence_field='primary', number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Secondary structure labels for a set of proteins determined by the local structures of protein residues in their natural state

Statistics:

• #Train: 8,678

• #Valid: 2,170

• #Test: 513

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• sequence_field (str, optional) – name of the field of protein sequence in lmdb files

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

ProteinNet

class ProteinNet(path, verbose=1, sequence_field='primary', number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

A set of proteins with 3D structures for the contact prediction task.

Statistics:

• #Train: 25,299

• #Valid: 224

• #Test: 40

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• sequence_field (str, optional) – name of the field of protein sequence in lmdb files

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the proteins are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Protein-Protein Prediction Datasets

HumanPPI

class HumanPPI(path, verbose=1, number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Binary labels indicating whether two human proteins interact or not.

Statistics:

• #Train: 6,844

• #Valid: 277

• #Test: 227

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the protein pairs are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

YeastPPI

class YeastPPI(path, verbose=1, number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Binary labels indicating whether two yeast proteins interact or not.

Statistics:

• #Train: 1,668

• #Valid: 131

• #Test: 373

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the protein pairs are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

PPIAffinity

class PPIAffinity(path, verbose=1, number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

The binding affinity values measured by \(p_{K_d}\) between two proteins.

Statistics:

• #Train: 2,127

• #Valid: 212

• #Test: 343

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the protein pairs are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Protein Ligand Prediction Datasets

BindingDB

class BindingDB(path, verbose=1, number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

The BindingDB dataset with binding affinity indicating the interaction strength between pairs of protein and ligand.

Statistics:

• #Train: 7,900

• #Valid: 878

• #Test: 5,230

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the protein-ligand pairs are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

PDBBind

class PDBBind(path, verbose=1, number_field='num_examples', transform=None, lazy=False, attributes=None, atom_feature='default', bond_feature='default', residue_feature='default', mol_feature=None, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

The PDBbind-2019 dataset with binding affinity indicating the interaction strength between pairs of protein and ligand.

Statistics:

• #Train: 16,436

• #Valid: 937

• #Test: 285

Parameters

• path (str) – the path to store the dataset

• verbose (int, optional) – output verbose level

• number_field (str, optional) – name of the field of sample count in lmdb files

• transform (Callable, optional) – protein sequence transformation function

• lazy (bool, optional) – if lazy mode is used, the protein-ligand pairs are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• attributes (dict of list) – protein-level attributes

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• residue_feature (str, list of str, optional) – residue features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

Retrosynthesis Datasets

USPTO50k

class USPTO50k(path, as_synthon=False, verbose=1, transform=None, lazy=False, atom_feature='default', bond_feature='default', mol_feature=None, with_hydrogen=False, kekulize=False, node_feature=None, edge_feature=None, graph_feature=None)

Chemical reactions extracted from USPTO patents.

Statistics:

• #Reaction: 50,017

• #Reaction class: 10

Parameters

• path (str) – path to store the dataset

• as_synthon (bool, optional) – whether decompose (reactant, product) pairs into (reactant, synthon) pairs

• verbose (int, optional) – output verbose level

• transform (Callable, optional) – data transformation function

• lazy (bool, optional) – if lazy mode is used, the molecules are processed in the dataloader. This may slow down the data loading process, but save a lot of CPU memory and dataset loading time.

• atom_feature (str or list of str, optional) – atom features to extract

• bond_feature (str or list of str, optional) – bond features to extract

• mol_feature (str or list of str, optional) – molecule features to extract

• with_hydrogen (bool, optional) – store hydrogens in the molecule graph. By default, hydrogens are dropped

• kekulize (bool, optional) – convert aromatic bonds to single/double bonds. Note this only affects the relation in edge_list. For bond_type, aromatic bonds are always stored explicitly. By default, aromatic bonds are stored.

• node_feature (str or list of str, optional) – deprecated alias of atom_feature

• edge_feature (str or list of str, optional) – deprecated alias of bond_feature

• graph_feature (str or list of str, optional) – deprecated alias of mol_feature

property reaction_types

All reaction types.

Citation Network Datasets

Cora

class Cora(path, verbose=1)

A citation network of scientific publications with binary word features.

Statistics:

• #Node: 2,708

• #Edge: 5,429

• #Class: 7

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

CiteSeer

class CiteSeer(path, verbose=1)

A citation network of scientific publications with binary word features.

Statistics:

• #Node: 3,327

• #Edge: 8,059

• #Class: 6

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level

PubMed

class PubMed(path, verbose=1)

A citation network of scientific publications with TF-IDF word features.

Statistics:

• #Node: 19,717

• #Edge: 44,338

• #Class: 3

Parameters

• path (str) – path to store the dataset

• verbose (int, optional) – output verbose level