Merge pull request #1156 from lilleswing/kb-atom-features (ff662561) · Commits · 钟慕尧 / deepchem

deepchem/data/datasets.py

+6 −6

Original line number	Diff line number	Diff line
		@@ -200,8 +200,8 @@ class Dataset(object):
		>>> dataset = NumpyDataset(np.ones((2,2)))
		>>> for x, y, w, id in dataset.itersamples():
		... print(x.tolist(), y.tolist(), w.tolist(), id)
		[1.0 1.0] [0.0] [0.0] 0
		[1.0 1.0] [0.0] [0.0] 1
		[1.0, 1.0] [0.0] [0.0] 0
		[1.0, 1.0] [0.0] [0.0] 1
		"""
		raise NotImplementedError()

		@@ -409,8 +409,8 @@ class NumpyDataset(Dataset):
		>>> dataset = NumpyDataset(np.ones((2,2)))
		>>> for x, y, w, id in dataset.itersamples():
		... print(x.tolist(), y.tolist(), w.tolist(), id)
		[1.0 1.0] [0.0] [0.0] 0
		[1.0 1.0] [0.0] [0.0] 1
		[1.0, 1.0] [0.0] [0.0] 0
		[1.0, 1.0] [0.0] [0.0] 1
		"""
		n_samples = self._X.shape[0]
		return ((self._X[i], self._y[i], self._w[i], self._ids[i])
		@@ -889,8 +889,8 @@ class DiskDataset(Dataset):
		>>> dataset = DiskDataset.from_numpy(np.ones((2,2)), np.ones((2,1)), verbose=False)
		>>> for x, y, w, id in dataset.itersamples():
		... print(x.tolist(), y.tolist(), w.tolist(), id)
		[1.0 1.0] [0.0] [0.0] 0
		[1.0 1.0] [0.0] [0.0] 1
		[1.0, 1.0] [1.0] [1.0] 0
		[1.0, 1.0] [1.0] [1.0] 1
		"""

		def iterate(dataset):

deepchem/feat/graph_features.py

+48 −5

Original line number	Diff line number	Diff line
		@@ -3,7 +3,6 @@ from __future__ import unicode_literals

		import numpy as np
		from rdkit import Chem
		import itertools, operator

		from deepchem.feat import Featurizer
		from deepchem.feat.mol_graphs import ConvMol, WeaveMol
		@@ -264,7 +263,8 @@ def find_distance(a1, num_atoms, canon_adj_list, max_distance=7):
		class ConvMolFeaturizer(Featurizer):
		name = ['conv_mol']

		def __init__(self, master_atom=False, use_chirality=False):
		def __init__(self, master_atom=False, use_chirality=False,
		atom_properties=[]):
		"""
		Parameters
		----------
		@@ -274,7 +274,20 @@ class ConvMolFeaturizer(Featurizer):
		the molecule. This technique is briefly discussed in
		Neural Message Passing for Quantum Chemistry
		https://arxiv.org/pdf/1704.01212.pdf

		use_chirality: Boolean
		if true then make the resulting atom features aware of the
		chirality of the molecules in question
		atom_properties: list of string or None
		properties in the RDKit Mol object to use as additional
		atom-level features in the larger molecular feature. If None,
		then no atom-level properties are used. Properties should be in the
		RDKit mol object should be in the form
		atom XXXXXXXX NAME
		where XXXXXXXX is a zero-padded 8 digit number coresponding to the
		zero-indexed atom index of each atom and NAME is the name of the property
		provided in atom_properties. So "atom 00000000 sasa" would be the
		name of the molecule level property in mol where the solvent
		accessible surface area of atom 0 would be stored.

		Since ConvMol is an object and not a numpy array, need to set dtype to
		object.
		@@ -282,12 +295,39 @@ class ConvMolFeaturizer(Featurizer):
		self.dtype = object
		self.master_atom = master_atom
		self.use_chirality = use_chirality
		self.atom_properties = list(atom_properties)

		def _get_atom_properties(self, atom):
		"""
		For a given input RDKit atom return the values of the properties
		requested when initializing the featurize. See the __init__ of the
		class for a full description of the names of the properties

		Parameters
		----------
		atom: RDKit.rdchem.Atom
		Atom to get the properties of
		returns a numpy lists of floats of the same size as self.atom_properties
		"""
		values = []
		for prop in self.atom_properties:
		mol_prop_name = str("atom %08d %s" % (atom.GetIdx(), prop))
		try:
		values.append(float(atom.GetOwningMol().GetProp(mol_prop_name)))
		except KeyError:
		raise KeyError("No property %s found in %s in %s" %
		(mol_prop_name, atom.GetOwningMol(), self))
		return np.array(values)

		def _featurize(self, mol):
		"""Encodes mol as a ConvMol object."""
		# Get the node features
		idx_nodes = [(a.GetIdx(), atom_features(
		a, use_chirality=self.use_chirality)) for a in mol.GetAtoms()]
		idx_nodes = [(a.GetIdx(),
		np.concatenate((atom_features(
		a, use_chirality=self.use_chirality),
		self._get_atom_properties(a))))
		for a in mol.GetAtoms()]

		idx_nodes.sort() # Sort by ind to ensure same order as rd_kit
		idx, nodes = list(zip(*idx_nodes))

		@@ -315,6 +355,9 @@ class ConvMolFeaturizer(Featurizer):

		return ConvMol(nodes, canon_adj_list)

		def feature_length(self):
		return 75 + len(self.atom_properties)


		class WeaveFeaturizer(Featurizer):
		name = ['weave_mol']

deepchem/models/tensorgraph/models/graph_models.py

+9 −3

Original line number	Diff line number	Diff line
		@@ -672,6 +672,7 @@ class GraphConvTensorGraph(TensorGraph):
		dense_layer_size=128,
		dropout=0.0,
		mode="classification",
		number_atom_features=75,
		**kwargs):
		"""
		Parameters
		@@ -686,6 +687,10 @@ class GraphConvTensorGraph(TensorGraph):
		Droupout dropout probability. Dropout is applied after the per Atom Level Dense Layer
		mode: str
		Either "classification" or "regression"
		number_atom_features: int
		75 is the default number of atom features created, but
		this can vary if various options are passed to the
		function atom_features in graph_features
		"""
		self.n_tasks = n_tasks
		self.mode = mode
		@@ -694,6 +699,7 @@ class GraphConvTensorGraph(TensorGraph):
		self.dropout = dropout
		self.graph_conv_layers = graph_conv_layers
		kwargs['use_queue'] = False
		self.number_atom_features = number_atom_features
		super(GraphConvTensorGraph, self).__init__(**kwargs)
		self.build_graph()

		@@ -701,7 +707,7 @@ class GraphConvTensorGraph(TensorGraph):
		"""
		Building graph structures:
		"""
		self.atom_features = Feature(shape=(None, 75))
		self.atom_features = Feature(shape=(None, self.number_atom_features))
		self.degree_slice = Feature(shape=(None, 2), dtype=tf.int32)
		self.membership = Feature(shape=(None,), dtype=tf.int32)

deepchem/models/tensorgraph/models/test_graph_models.py

+34 −0

Original line number	Diff line number	Diff line
		@@ -9,6 +9,7 @@ from deepchem.models import TensorGraph
		from deepchem.molnet.load_function.delaney_datasets import load_delaney
		from deepchem.models.tensorgraph.layers import ReduceSum, L2Loss
		from deepchem.models import WeaveTensorGraph
		from deepchem.feat import ConvMolFeaturizer


		class TestGraphModels(unittest.TestCase):
		@@ -82,6 +83,39 @@ class TestGraphModels(unittest.TestCase):
		assert mu.shape == (len(dataset), len(tasks))
		assert sigma.shape == (len(dataset), len(tasks))

		def test_graph_conv_atom_features(self):
		tasks, dataset, transformers, metric = self.get_dataset(
		'regression', 'Raw', num_tasks=1)

		atom_feature_name = 'feature'
		y = []
		for mol in dataset.X:
		atom_features = []
		for atom in mol.GetAtoms():
		val = np.random.normal()
		mol.SetProp("atom %08d %s" % (atom.GetIdx(), atom_feature_name),
		str(val))
		atom_features.append(np.random.normal())
		y.append(np.sum(atom_features))

		featurizer = ConvMolFeaturizer(atom_properties=[atom_feature_name])
		X = featurizer.featurize(dataset.X)
		dataset = deepchem.data.NumpyDataset(X, np.array(y))
		batch_size = 50
		model = GraphConvTensorGraph(
		len(tasks),
		number_atom_features=featurizer.feature_length(),
		batch_size=batch_size,
		mode='regression')

		model.fit(dataset, nb_epoch=1)
		y_pred1 = model.predict(dataset)
		model.save()

		model2 = TensorGraph.load_from_dir(model.model_dir)
		y_pred2 = model2.predict(dataset)
		self.assertTrue(np.all(y_pred1 == y_pred2))

		def test_change_loss_function(self):
		tasks, dataset, transformers, metric = self.get_dataset(
		'regression', 'GraphConv', num_tasks=1)

.travis.yml

+1 −1

File changed.

Contains only whitespace changes.

Admin message