[DGL-LifeSci] WLN for Reaction Center Prediction (#1360)

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

* Update

apps/life_sci/setup.py

@@ -26,14 +26,15 @@ setup(
     packages=[package for package in find_packages()
               if package.startswith('dgllife')],
     install_requires=[
-        'dgl>=0.4',
         'torch>=1'
         'scikit-learn>=0.21.2',
         'pandas>=0.25.1',
-        'requests>=2.22.0'
+        'requests>=2.22.0',
+        'tqdm'
     ],
     url='https://github.com/dmlc/dgl/tree/master/apps/life_sci',
     classifiers=[
+        'Development Status :: 3 - Alpha',
         'Programming Language :: Python :: 3',
-    ]
+    ],
 )

apps/life_sci/tests/data/test_datasets.py

 import os
 
 from dgllife.data import *
+from dgllife.data.uspto import get_bond_changes, process_file
 
 def remove_file(fname):
     if os.path.isfile(fname):
@@ -10,14 +11,17 @@ def remove_file(fname):
             pass
 
 def test_pubchem_aromaticity():
+    print('Test pubchem aromaticity')
     dataset = PubChemBioAssayAromaticity()
     remove_file('pubchem_aromaticity_dglgraph.bin')
 
 def test_tox21():
+    print('Test Tox21')
     dataset = Tox21()
     remove_file('tox21_dglgraph.bin')
 
 def test_alchemy():
+    print('Test Alchemy')
     dataset = TencentAlchemyDataset(mode='valid',
                                     node_featurizer=None,
                                     edge_featurizer=None)
@@ -27,10 +31,47 @@ def test_alchemy():
                                     load=False)
 
 def test_pdbbind():
+    print('Test PDBBind')
     dataset = PDBBind(subset='core', remove_hs=True)
 
+def test_wln_reaction():
+    print('Test datasets for reaction prediction with WLN')
+
+    reaction1 = '[CH2:15]([CH:16]([CH3:17])[CH3:18])[Mg+:19].[CH2:20]1[O:21][CH2:22][CH2:23]' \
+                '[CH2:24]1.[Cl-:14].[OH:1][c:2]1[n:3][cH:4][c:5]([C:6](=[O:7])[N:8]([O:9]' \
+                '[CH3:10])[CH3:11])[cH:12][cH:13]1>>[OH:1][c:2]1[n:3][cH:4][c:5]([C:6](=[O:7])' \
+                '[CH2:15][CH:16]([CH3:17])[CH3:18])[cH:12][cH:13]1\n'
+    reaction2 = '[CH3:14][NH2:15].[N+:1](=[O:2])([O-:3])[c:4]1[cH:5][c:6]([C:7](=[O:8])[OH:9])' \
+                '[cH:10][cH:11][c:12]1[Cl:13].[OH2:16]>>[N+:1](=[O:2])([O-:3])[c:4]1[cH:5][c:6]' \
+                '([C:7](=[O:8])[OH:9])[cH:10][cH:11][c:12]1[NH:15][CH3:14]\n'
+    reactions = [reaction1, reaction2]
+
+    # Test utility functions
+    assert get_bond_changes(reaction2) == {('12', '13', 0.0), ('12', '15', 1.0)}
+    with open('test.txt', 'w') as f:
+        for reac in reactions:
+            f.write(reac)
+    process_file('test.txt')
+    with open('test.txt.proc', 'r') as f:
+        lines = f.readlines()
+        for i in range(len(lines)):
+            l = lines[i].strip()
+            react = reactions[i].strip()
+            bond_changes = get_bond_changes(react)
+            assert l == '{} {}'.format(
+                react,
+                ';'.join(['{}-{}-{}'.format(x[0], x[1], x[2]) for x in bond_changes]))
+    remove_file('test.txt.proc')
+
+    # Test configured dataset
+    dataset = WLNReactionDataset('test.txt', 'test_graphs.bin')
+    remove_file('test.txt')
+    remove_file('test.txt.proc')
+    remove_file('test_graphs.bin')
+
 if __name__ == '__main__':
     test_pubchem_aromaticity()
     test_tox21()
     test_alchemy()
     test_pdbbind()
+    test_wln_reaction()

apps/life_sci/tests/model/test_gnn.py

@@ -209,6 +209,32 @@ def test_mpnn_gnn():
     assert gnn(g, node_feats, edge_feats).shape == torch.Size([3, 2])
     assert gnn(bg, batch_node_feats, batch_edge_feats).shape == torch.Size([8, 2])
 
+def test_wln():
+    if torch.cuda.is_available():
+        device = torch.device('cuda:0')
+    else:
+        device = torch.device('cpu')
+
+    g, node_feats, edge_feats = test_graph3()
+    g, node_feats, edge_feats = g.to(device), node_feats.to(device), edge_feats.to(device)
+    bg, batch_node_feats, batch_edge_feats = test_graph4()
+    bg, batch_node_feats, batch_edge_feats = bg.to(device), batch_node_feats.to(device), \
+                                             batch_edge_feats.to(device)
+
+    # Test default setting
+    gnn = WLN(node_in_feats=1,
+              edge_in_feats=2)
+    assert gnn(g, node_feats, edge_feats).shape == torch.Size([3, 300])
+    assert gnn(bg, batch_node_feats, batch_edge_feats).shape == torch.Size([8, 300])
+
+    # Test configured setting
+    gnn = WLN(node_in_feats=1,
+              edge_in_feats=2,
+              node_out_feats=3,
+              n_layers=1)
+    assert gnn(g, node_feats, edge_feats).shape == torch.Size([3, 3])
+    assert gnn(bg, batch_node_feats, batch_edge_feats).shape == torch.Size([8, 3])
+
 if __name__ == '__main__':
     test_gcn()
     test_gat()
@@ -216,3 +242,4 @@ if __name__ == '__main__':
     test_schnet_gnn()
     test_mgcn_gnn()
     test_mpnn_gnn()
+    test_wln()

apps/life_sci/tests/model/test_reaction_prediction.py

+import dgl
+import torch
+
+from dgl import DGLGraph
+
+from dgllife.model.model_zoo import *
+
+def get_complete_graph(num_nodes):
+    edge_list = []
+    for i in range(num_nodes):
+        for j in range(num_nodes):
+            edge_list.append((i, j))
+    return DGLGraph(edge_list)
+
+def test_graph1():
+    """
+    Bi-directed graphs and complete graphs for the molecules.
+    In addition to node features/edge features, we also return
+    features for the pairs of nodes.
+    """
+    mol_graph = DGLGraph([(0, 1), (0, 2), (1, 2)])
+    node_feats = torch.arange(mol_graph.number_of_nodes()).float().reshape(-1, 1)
+    edge_feats = torch.arange(2 * mol_graph.number_of_edges()).float().reshape(-1, 2)
+
+    complete_graph = get_complete_graph(mol_graph.number_of_nodes())
+    atom_pair_feats = torch.arange(complete_graph.number_of_edges()).float().reshape(-1, 1)
+
+    return mol_graph, node_feats, edge_feats, complete_graph, atom_pair_feats
+
+def test_graph2():
+    """Batched version of test_graph1"""
+    mol_graph1 = DGLGraph([(0, 1), (0, 2), (1, 2)])
+    mol_graph2 = DGLGraph([(0, 1), (1, 2), (1, 3), (1, 4)])
+    batch_mol_graph = dgl.batch([mol_graph1, mol_graph2])
+    node_feats = torch.arange(batch_mol_graph.number_of_nodes()).float().reshape(-1, 1)
+    edge_feats = torch.arange(2 * batch_mol_graph.number_of_edges()).float().reshape(-1, 2)
+
+    complete_graph1 = get_complete_graph(mol_graph1.number_of_nodes())
+    complete_graph2 = get_complete_graph(mol_graph2.number_of_nodes())
+    batch_complete_graph = dgl.batch([complete_graph1, complete_graph2])
+    atom_pair_feats = torch.arange(batch_complete_graph.number_of_edges()).float().reshape(-1, 1)
+
+    return batch_mol_graph, node_feats, edge_feats, batch_complete_graph, atom_pair_feats
+
+def test_wln_reaction_center():
+    if torch.cuda.is_available():
+        device = torch.device('cuda:0')
+    else:
+        device = torch.device('cpu')
+
+    mol_graph, node_feats, edge_feats, complete_graph, atom_pair_feats = test_graph1()
+    mol_graph = mol_graph.to(device)
+    node_feats, edge_feats = node_feats.to(device), edge_feats.to(device)
+    complete_graph = complete_graph.to(device)
+    atom_pair_feats = atom_pair_feats.to(device)
+
+    batch_mol_graph, batch_node_feats, batch_edge_feats, batch_complete_graph, \
+    batch_atom_pair_feats = test_graph2()
+    batch_mol_graph = batch_mol_graph.to(device)
+    batch_node_feats, batch_edge_feats = batch_node_feats.to(device), batch_edge_feats.to(device)
+    batch_complete_graph = batch_complete_graph.to(device)
+    batch_atom_pair_feats = batch_atom_pair_feats.to(device)
+
+    # Test default setting
+    model = WLNReactionCenter(node_in_feats=1,
+                              edge_in_feats=2,
+                              node_pair_in_feats=1).to(device)
+    assert model(mol_graph, complete_graph, node_feats, edge_feats, atom_pair_feats).shape == \
+           torch.Size([complete_graph.number_of_edges(), 5])
+    assert model(batch_mol_graph, batch_complete_graph, batch_node_feats,
+                 batch_edge_feats, batch_atom_pair_feats).shape == \
+           torch.Size([batch_complete_graph.number_of_edges(), 5])
+
+    # Test configured setting
+    model = WLNReactionCenter(node_in_feats=1,
+                              edge_in_feats=2,
+                              node_pair_in_feats=1,
+                              node_out_feats=1,
+                              n_layers=1,
+                              n_tasks=1).to(device)
+    assert model(mol_graph, complete_graph, node_feats, edge_feats, atom_pair_feats).shape == \
+           torch.Size([complete_graph.number_of_edges(), 1])
+    assert model(batch_mol_graph, batch_complete_graph, batch_node_feats,
+                 batch_edge_feats, batch_atom_pair_feats).shape == \
+           torch.Size([batch_complete_graph.number_of_edges(), 1])
+
+if __name__ == '__main__':
+    test_wln_reaction_center()

apps/life_sci/tests/utils/test_featurizers.py

@@ -54,6 +54,16 @@ def test_atom_total_degree():
     assert atom_total_degree(mol.GetAtomWithIdx(0)) == [4]
     assert atom_total_degree(mol.GetAtomWithIdx(2)) == [2]
 
+def test_atom_explicit_valence_one_hot():
+    mol = test_mol1()
+    assert atom_implicit_valence_one_hot(mol.GetAtomWithIdx(0), [1, 2, 3]) == [1, 0, 0]
+    assert atom_implicit_valence_one_hot(mol.GetAtomWithIdx(1), [1, 2, 3]) == [0, 1, 0]
+
+def test_atom_explicit_valence():
+    mol = test_mol1()
+    assert atom_explicit_valence(mol.GetAtomWithIdx(0)) == [1]
+    assert atom_explicit_valence(mol.GetAtomWithIdx(1)) == [2]
+
 def test_atom_implicit_valence_one_hot():
     mol = test_mol1()
     assert atom_implicit_valence_one_hot(mol.GetAtomWithIdx(0), [1, 2, 3]) == [0, 0, 1]
@@ -106,6 +116,18 @@ def test_atom_is_aromatic():
     mol = test_mol2()
     assert atom_is_aromatic(mol.GetAtomWithIdx(0)) == [1]
 
+def test_atom_is_in_ring_one_hot():
+    mol = test_mol1()
+    assert atom_is_in_ring_one_hot(mol.GetAtomWithIdx(0)) == [1, 0]
+    mol = test_mol2()
+    assert atom_is_in_ring_one_hot(mol.GetAtomWithIdx(0)) == [0, 1]
+
+def test_atom_is_in_ring():
+    mol = test_mol1()
+    assert atom_is_in_ring(mol.GetAtomWithIdx(0)) == [0]
+    mol = test_mol2()
+    assert atom_is_in_ring(mol.GetAtomWithIdx(0)) == [1]
+
 def test_atom_chiral_tag_one_hot():
     mol = test_mol1()
     assert atom_chiral_tag_one_hot(mol.GetAtomWithIdx(0)) == [1, 0, 0, 0]
@@ -246,6 +268,7 @@ if __name__ == '__main__':
     test_atom_degree()
     test_atom_total_degree_one_hot()
     test_atom_total_degree()
+    test_atom_explicit_valence()
     test_atom_implicit_valence_one_hot()
     test_atom_implicit_valence()
     test_atom_hybridization_one_hot()
@@ -257,6 +280,8 @@ if __name__ == '__main__':
     test_atom_num_radical_electrons()
     test_atom_is_aromatic_one_hot()
     test_atom_is_aromatic()
+    test_atom_is_in_ring_one_hot()
+    test_atom_is_in_ring()
     test_atom_chiral_tag_one_hot()
     test_atom_mass()
     test_concat_featurizer()

apps/life_sci/tests/utils/test_mol_to_graph.py

@@ -7,6 +7,9 @@ from rdkit import Chem
 
 test_smiles1 = 'CCO'
 test_smiles2 = 'Fc1ccccc1'
+test_smiles3 = '[CH2:1]([CH3:2])[N:3]1[CH2:4][CH2:5][C:6]([CH3:16])' \
+               '([CH3:17])[c:7]2[cH:8][cH:9][c:10]([N+:13]([O-:14])=[O:15])' \
+               '[cH:11][c:12]21.[CH3:18][CH2:19][O:20][C:21]([CH3:22])=[O:23]'
 
 class TestAtomFeaturizer(BaseAtomFeaturizer):
     def __init__(self):
@@ -37,6 +40,16 @@ def test_smiles_to_bigraph():
                                                         [1.], [1.], [1.], [1.], [1.],
                                                         [1.], [1.], [1.], [1.]]))
 
+    # Test the case where atoms come with a default order and we do not
+    # want to change the order, which is related to the application of
+    # reaction center prediction.
+    g3 = smiles_to_bigraph(test_smiles3, node_featurizer=test_node_featurizer,
+                           canonical_atom_order=False)
+    assert torch.allclose(g3.ndata['hv'], torch.tensor([[6.], [6.], [7.], [6.], [6.], [6.],
+                                                        [6.], [6.], [6.], [6.], [6.], [6.],
+                                                        [7.], [8.], [8.], [6.], [6.], [6.],
+                                                        [6.], [8.], [6.], [6.], [8.]]))
+
 def test_mol_to_bigraph():
     mol1 = Chem.MolFromSmiles(test_smiles1)
     g1 = mol_to_bigraph(mol1, add_self_loop=True)
@@ -57,24 +70,56 @@ def test_mol_to_bigraph():
                                                         [1.], [1.], [1.], [1.], [1.],
                                                         [1.], [1.], [1.], [1.]]))
 
+    # Test the case where atoms come with a default order and we do not
+    # want to change the order, which is related to the application of
+    # reaction center prediction.
+    mol3 = Chem.MolFromSmiles(test_smiles3)
+    g3 = mol_to_bigraph(mol3, node_featurizer=test_node_featurizer,
+                        canonical_atom_order=False)
+    assert torch.allclose(g3.ndata['hv'], torch.tensor([[6.], [6.], [7.], [6.], [6.], [6.],
+                                                        [6.], [6.], [6.], [6.], [6.], [6.],
+                                                        [7.], [8.], [8.], [6.], [6.], [6.],
+                                                        [6.], [8.], [6.], [6.], [8.]]))
+
 def test_smiles_to_complete_graph():
     test_node_featurizer = TestAtomFeaturizer()
-    g = smiles_to_complete_graph(test_smiles1, add_self_loop=False,
+    g1 = smiles_to_complete_graph(test_smiles1, add_self_loop=False,
                                  node_featurizer=test_node_featurizer)
-    src, dst = g.edges()
+    src, dst = g1.edges()
     assert torch.allclose(src, torch.LongTensor([0, 0, 1, 1, 2, 2]))
     assert torch.allclose(dst, torch.LongTensor([1, 2, 0, 2, 0, 1]))
-    assert torch.allclose(g.ndata['hv'], torch.tensor([[6.], [8.], [6.]]))
+    assert torch.allclose(g1.ndata['hv'], torch.tensor([[6.], [8.], [6.]]))
+
+    # Test the case where atoms come with a default order and we do not
+    # want to change the order, which is related to the application of
+    # reaction center prediction.
+    g2 = smiles_to_complete_graph(test_smiles3, node_featurizer=test_node_featurizer,
+                                  canonical_atom_order=False)
+    assert torch.allclose(g2.ndata['hv'], torch.tensor([[6.], [6.], [7.], [6.], [6.], [6.],
+                                                        [6.], [6.], [6.], [6.], [6.], [6.],
+                                                        [7.], [8.], [8.], [6.], [6.], [6.],
+                                                        [6.], [8.], [6.], [6.], [8.]]))
 
 def test_mol_to_complete_graph():
     test_node_featurizer = TestAtomFeaturizer()
     mol1 = Chem.MolFromSmiles(test_smiles1)
-    g = mol_to_complete_graph(mol1, add_self_loop=False,
+    g1 = mol_to_complete_graph(mol1, add_self_loop=False,
                                node_featurizer=test_node_featurizer)
-    src, dst = g.edges()
+    src, dst = g1.edges()
     assert torch.allclose(src, torch.LongTensor([0, 0, 1, 1, 2, 2]))
     assert torch.allclose(dst, torch.LongTensor([1, 2, 0, 2, 0, 1]))
-    assert torch.allclose(g.ndata['hv'], torch.tensor([[6.], [8.], [6.]]))
+    assert torch.allclose(g1.ndata['hv'], torch.tensor([[6.], [8.], [6.]]))
+
+    # Test the case where atoms come with a default order and we do not
+    # want to change the order, which is related to the application of
+    # reaction center prediction.
+    mol2 = Chem.MolFromSmiles(test_smiles3)
+    g2 = mol_to_complete_graph(mol2, node_featurizer=test_node_featurizer,
+                               canonical_atom_order=False)
+    assert torch.allclose(g2.ndata['hv'], torch.tensor([[6.], [6.], [7.], [6.], [6.], [6.],
+                                                        [6.], [6.], [6.], [6.], [6.], [6.],
+                                                        [7.], [8.], [8.], [6.], [6.], [6.],
+                                                        [6.], [8.], [6.], [6.], [8.]]))
 
 def test_k_nearest_neighbors():
     coordinates = np.array([[0.1, 0.1, 0.1],