Updates to notebooks

%% Cell type:markdown id: tags:

# Graph Convolutions For Tox21

In this notebook, we will explore the use of TensorGraph to create graph convolutional models with DeepChem. In particular, we will build a graph convolutional network on the Tox21 dataset.

In this notebook, we will explore the use of KerasModel to create graph convolutional models with DeepChem. In particular, we will build a graph convolutional network on the Tox21 dataset.

Let's start with some basic imports.

%% Cell type:code id: tags:

``` python

from __future__ import division

from __future__ import print_function

from __future__ import unicode_literals

import numpy as np

import tensorflow as tf

import deepchem as dc

from deepchem.models.graph_models import GraphConvModel

```

%% Cell type:markdown id: tags:

Now, let's use MoleculeNet to load the Tox21 dataset. We need to make sure to process the data in a way that graph convolutional networks can use For that, we make sure to set the featurizer option to 'GraphConv'. The MoleculeNet call will return a training set, an validation set, and a test set for us to use. The call also returns `transformers`, a list of data transformations that were applied to preprocess the dataset. (Most deep networks are quite finicky and require a set of data transformations to ensure that training proceeds stably.)

%% Cell type:code id: tags:

``` python

# Load Tox21 dataset

tox21_tasks, tox21_datasets, transformers = dc.molnet.load_tox21(featurizer='GraphConv')

train_dataset, valid_dataset, test_dataset = tox21_datasets

```

%% Output

    Loading dataset from disk.

    Loading dataset from disk.

    Loading dataset from disk.

%% Cell type:markdown id: tags:

Let's now train a graph convolutional network on this dataset. DeepChem has the class `GraphConvModel` that wraps a standard graph convolutional architecture underneath the hood for user convenience. Let's instantiate an object of this class and train it on our dataset.

%% Cell type:code id: tags:

``` python

model = GraphConvModel(

    len(tox21_tasks), batch_size=50, mode='classification')

n_tasks = len(tox21_tasks)

model = GraphConvModel(n_tasks, batch_size=50, mode='classification')

# Set nb_epoch=10 for better results.

model.fit(train_dataset, nb_epoch=1)

```

%% Output

    /home/leswing/miniconda3/envs/deepchem/lib/python3.5/site-packages/tensorflow/python/ops/gradients_impl.py:95: UserWarning: Converting sparse IndexedSlices to a dense Tensor of unknown shape. This may consume a large amount of memory.

      "Converting sparse IndexedSlices to a dense Tensor of unknown shape. "

    Starting epoch 0

    Ending global_step 126: Average loss 590.739

    TIMING: model fitting took 7.317 s

    590.7391258118645

    0.8578314024668473

%% Cell type:markdown id: tags:

Let's try to evaluate the performance of the model we've trained. For this, we need to define a metric, a measure of model performance. `dc.metrics` holds a collection of metrics already. For this dataset, it is standard to use the ROC-AUC score, the area under the receiver operating characteristic curve (which measures the tradeoff between precision and recall). Luckily, the ROC-AUC score is already available in DeepChem.

To measure the performance of the model under this metric, we can use the convenience function `model.evaluate()`.

%% Cell type:code id: tags:

``` python

metric = dc.metrics.Metric(

    dc.metrics.roc_auc_score, np.mean, mode="classification")

metric = dc.metrics.Metric(dc.metrics.roc_auc_score, np.mean)

print("Evaluating model")

train_scores = model.evaluate(train_dataset, [metric], transformers)

print("Training ROC-AUC Score: %f" % train_scores["mean-roc_auc_score"])

valid_scores = model.evaluate(valid_dataset, [metric], transformers)

print("Validation ROC-AUC Score: %f" % valid_scores["mean-roc_auc_score"])

```

%% Output

    Evaluating model

    computed_metrics: [0.80045699830862893, 0.83618637604367374, 0.83908539936708681, 0.77873855933094183, 0.67692252993044244, 0.75578036941489168, 0.75895796821704797, 0.70234314980793855, 0.76387081283102387, 0.65924917162534913, 0.78448201448364341, 0.76675448900822296]

    Training ROC-AUC Score: 0.760236

    computed_metrics: [0.71533171721169553, 0.74090608465608465, 0.81106357802757933, 0.70627859684799188, 0.63177272727272715, 0.6326016835811501, 0.61491865697473169, 0.71286314850043442, 0.67676006592889104, 0.51656328658755846, 0.75414979999520937, 0.6603359173126615]

    Validation ROC-AUC Score: 0.681129

%% Cell type:markdown id: tags:

What's going on under the hood? Could we build `GraphConvModel` ourselves? Of course! The first step is to create a `TensorGraph` object. This object will hold the "computational graph" that defines the computation that a graph convolutional network will perform.

%% Cell type:code id: tags:

``` python

from deepchem.models.tensorgraph.tensor_graph import TensorGraph

tg = TensorGraph(use_queue=False)

```

    computed_metrics: [0.7846645298316683, 0.8255705081888247, 0.8389671176952215, 0.8060793455430841, 0.6711073801951157, 0.7740920978757578, 0.7495887814740334, 0.7160807746060438, 0.767546874021283, 0.7088871155213989, 0.7977418645237055, 0.7459137439167645]

    Training ROC-AUC Score: 0.765520

    computed_metrics: [0.6741700885568693, 0.7308201058201058, 0.833249897601602, 0.7730045754956787, 0.6146818181818182, 0.6578554008339234, 0.6388888888888888, 0.7393550811432866, 0.7231457499411349, 0.5883899676375405, 0.7528563558408583, 0.6820844099913868]

    Validation ROC-AUC Score: 0.700709

%% Cell type:markdown id: tags:

Let's now define the inputs to our model. Conceptually, graph convolutions just requires a the structure of the molecule in question and a vector of features for every atom that describes the local chemical environment. However in practice, due to TensorFlow's limitations as a general programming environment, we have to have some auxiliary information as well preprocessed.

What's going on under the hood? Could we build `GraphConvModel` ourselves? Of course! The first step is to define the inputs to our model. Conceptually, graph convolutions just require the structure of the molecule in question and a vector of features for every atom that describes the local chemical environment. However in practice, due to TensorFlow's limitations as a general programming environment, we have to have some auxiliary information as well preprocessed.

`atom_features` holds a feature vector of length 75 for each atom. The other feature inputs are required to support minibatching in TensorFlow. `degree_slice` is an indexing convenience that makes it easy to locate atoms from all molecules with a given degree. `membership` determines the membership of atoms in molecules (atom `i` belongs to molecule `membership[i]`). `deg_adjs` is a list that contains adjacency lists grouped by atom degree For more details, check out the [code](https://github.com/deepchem/deepchem/blob/master/deepchem/feat/mol_graphs.py).

`atom_features` holds a feature vector of length 75 for each atom. The other inputs are required to support minibatching in TensorFlow. `degree_slice` is an indexing convenience that makes it easy to locate atoms from all molecules with a given degree. `membership` determines the membership of atoms in molecules (atom `i` belongs to molecule `membership[i]`). `deg_adjs` is a list that contains adjacency lists grouped by atom degree. For more details, check out the [code](https://github.com/deepchem/deepchem/blob/master/deepchem/feat/mol_graphs.py).

To define feature inputs in `TensorGraph`, we use the `Feature` layer. Conceptually, a `TensorGraph` is a mathematical graph composed of layer objects. `Features` layers have to be the root nodes of the graph since they consitute inputs.

To define feature inputs with Keras, we use the `Input` layer. Conceptually, a model is a mathematical graph composed of layer objects. `Input` layers have to be the root nodes of the graph since they consitute inputs.

%% Cell type:code id: tags:

``` python

from deepchem.models.tensorgraph.layers import Feature

import tensorflow as tf

from tensorflow.keras.layers import Input

atom_features = Feature(shape=(None, 75))

degree_slice = Feature(shape=(None, 2), dtype=tf.int32)

membership = Feature(shape=(None,), dtype=tf.int32)

atom_features = Input(shape=(75,))

degree_slice = Input(shape=(2,), dtype=tf.int32)

membership = Input(shape=tuple(), dtype=tf.int32)

deg_adjs = []

for i in range(0, 10 + 1):

    deg_adj = Feature(shape=(None, i + 1), dtype=tf.int32)

    deg_adj = Input(shape=(i+1,), dtype=tf.int32)

    deg_adjs.append(deg_adj)

```

%% Cell type:markdown id: tags:

Let's now implement the body of the graph convolutional network. `TensorGraph` has a number of layers that encode various graph operations. Namely, the `GraphConv`, `GraphPool` and `GraphGather` layers. We will also apply standard neural network layers such as `Dense` and `BatchNorm`.

Let's now implement the body of the graph convolutional network. DeepChem has a number of layers that encode various graph operations. Namely, the `GraphConv`, `GraphPool` and `GraphGather` layers. We will also apply standard neural network layers such as `Dense` and `BatchNormalization`.

The layers we're adding effect a "feature transformation" that will create one vector for each molecule.

%% Cell type:code id: tags:

``` python

from deepchem.models.tensorgraph.layers import Dense, GraphConv, BatchNorm

from deepchem.models.tensorgraph.layers import GraphPool, GraphGather

from tensorflow.keras.layers import Dense, BatchNormalization, Reshape, Softmax

from deepchem.models.layers import GraphConv, GraphPool, GraphGather

batch_size = 50

gc1 = GraphConv(

    64,

    activation_fn=tf.nn.relu,

    in_layers=[atom_features, degree_slice, membership] + deg_adjs)

batch_norm1 = BatchNorm(in_layers=[gc1])

gp1 = GraphPool(in_layers=[batch_norm1, degree_slice, membership] + deg_adjs)

gc2 = GraphConv(

    64,

    activation_fn=tf.nn.relu,

    in_layers=[gp1, degree_slice, membership] + deg_adjs)

batch_norm2 = BatchNorm(in_layers=[gc2])

gp2 = GraphPool(in_layers=[batch_norm2, degree_slice, membership] + deg_adjs)

dense = Dense(out_channels=128, activation_fn=tf.nn.relu, in_layers=[gp2])

batch_norm3 = BatchNorm(in_layers=[dense])

readout = GraphGather(

    batch_size=batch_size,

    activation_fn=tf.nn.tanh,

    in_layers=[batch_norm3, degree_slice, membership] + deg_adjs)

gc1 = GraphConv(64, activation_fn=tf.nn.relu)([atom_features, degree_slice, membership] + deg_adjs)

batch_norm1 = BatchNormalization()(gc1)

gp1 = GraphPool()([batch_norm1, degree_slice, membership] + deg_adjs)

gc2 = GraphConv(64, activation_fn=tf.nn.relu)([gp1, degree_slice, membership] + deg_adjs)

batch_norm2 = BatchNormalization()(gc2)

gp2 = GraphPool()([batch_norm2, degree_slice, membership] + deg_adjs)

dense = Dense(128, activation=tf.nn.relu)(gp2)

batch_norm3 = BatchNormalization()(dense)

readout = GraphGather(batch_size=batch_size, activation_fn=tf.nn.tanh)([batch_norm3, degree_slice, membership] + deg_adjs)

logits = Reshape((n_tasks, 2))(Dense(n_tasks*2)(readout))

softmax = Softmax()(logits)

```

%% Cell type:markdown id: tags:

Let's now make predictions from the `TensorGraph` model. Tox21 is a multitask dataset. That is, there are 12 different datasets grouped together, which share many common molecules, but with different outputs for each. As a result, we have to add a separate output layer for each task. We will use a `for` loop over the `tox21_tasks` list to make this happen. We need to add labels for each

We also have to define a loss for the model which tells the network the objective to minimize during training.

We have to tell `TensorGraph` which layers are outputs with `TensorGraph.add_output(layer)`. Similarly, we tell the network its loss with `TensorGraph.set_loss(loss)`.

Let's now create the `KerasModel`. To do that we specify the inputs and outputs to the model. We also have to define a loss for the model which tells the network the objective to minimize during training.

%% Cell type:code id: tags:

``` python

from deepchem.models.tensorgraph.layers import Dense, SoftMax, \

    SoftMaxCrossEntropy, WeightedError, Stack

from deepchem.models.tensorgraph.layers import Label, Weights

costs = []

labels = []

for task in range(len(tox21_tasks)):

    classification = Dense(

        out_channels=2, activation_fn=None, in_layers=[readout])

    softmax = SoftMax(in_layers=[classification])

    tg.add_output(softmax)

    label = Label(shape=(None, 2))

    labels.append(label)

    cost = SoftMaxCrossEntropy(in_layers=[label, classification])

    costs.append(cost)

all_cost = Stack(in_layers=costs, axis=1)

weights = Weights(shape=(None, len(tox21_tasks)))

loss = WeightedError(in_layers=[all_cost, weights])

tg.set_loss(loss)

inputs = [atom_features, degree_slice, membership] + deg_adjs

outputs = [softmax]

keras_model = tf.keras.Model(inputs=inputs, outputs=outputs)

loss = dc.models.losses.CategoricalCrossEntropy()

model = dc.models.KerasModel(keras_model, loss=loss)

```

%% Cell type:markdown id: tags:

Now that we've successfully defined our graph convolutional model in `TensorGraph`, we need to train it. We can call `fit()`, but we need to make sure that each minibatch of data populates all four `Feature` objects that we've created. For this, we need to create a Python generator that given a batch of data generates a dictionary whose keys are the `Feature` layers and whose values are Numpy arrays we'd like to use for this step of training.

Now that we've successfully defined our graph convolutional model, we need to train it. We can call `fit()`, but we need to make sure that each minibatch of data populates all the `Input` objects that we've created. For this, we need to create a Python generator that given a batch of data generates the lists of inputs, labels, and weights whose values are Numpy arrays we'd like to use for this step of training.

%% Cell type:code id: tags:

``` python

from deepchem.metrics import to_one_hot

from deepchem.feat.mol_graphs import ConvMol

def data_generator(dataset, epochs=1, predict=False, pad_batches=True):

  for epoch in range(epochs):

    if not predict:

        print('Starting epoch %i' % epoch)

    for ind, (X_b, y_b, w_b, ids_b) in enumerate(

        dataset.iterbatches(

            batch_size, pad_batches=pad_batches, deterministic=True)):

      d = {}

      for index, label in enumerate(labels):

        d[label] = to_one_hot(y_b[:, index])

      d[weights] = w_b

      multiConvMol = ConvMol.agglomerate_mols(X_b)

      d[atom_features] = multiConvMol.get_atom_features()

      d[degree_slice] = multiConvMol.deg_slice

      d[membership] = multiConvMol.membership

      inputs = [multiConvMol.get_atom_features(), multiConvMol.deg_slice, np.array(multiConvMol.membership)]

      for i in range(1, len(multiConvMol.get_deg_adjacency_lists())):

        d[deg_adjs[i - 1]] = multiConvMol.get_deg_adjacency_lists()[i]

      yield d

        inputs.append(multiConvMol.get_deg_adjacency_lists()[i])

      labels = [to_one_hot(y_b.flatten(), 2).reshape(-1, n_tasks, 2)]

      weights = [w_b]

      yield (inputs, labels, weights)

```

%% Cell type:markdown id: tags:

Now, we can train the model using `TensorGraph.fit_generator(generator)` which will use the generator we've defined to train the model.

Now, we can train the model using `KerasModel.fit_generator(generator)` which will use the generator we've defined to train the model.

%% Cell type:code id: tags:

``` python

# Epochs set to 1 to render tutorials online.

# Set epochs=10 for better results.

tg.fit_generator(data_generator(train_dataset, epochs=1))

model.fit_generator(data_generator(train_dataset, epochs=1))

```

%% Output

    Starting epoch 0

    Ending global_step 251: Average loss 530.84

    TIMING: model fitting took 6.949 s

    530.8396410260882

    0.888308482674452

%% Cell type:markdown id: tags:

Now that we have trained our graph convolutional method, let's evaluate its performance. We again have to use our defined generator to evaluate model performance.

%% Cell type:code id: tags:

``` python

metric = dc.metrics.Metric(

    dc.metrics.roc_auc_score, np.mean, mode="classification")

metric = dc.metrics.Metric(dc.metrics.roc_auc_score, np.mean)

def reshape_y_pred(y_true, y_pred):

    """

    TensorGraph.Predict returns a list of arrays, one for each output

    We also have to remove the padding on the last batch

    Metrics taks results of shape (samples, n_task, prob_of_class)

    GraphConv always pads batches, so we need to remove the predictions

    for the padding samples.  Also, it outputs two values for each task

    (probabilities of positive and negative), but we only want the positive

    probability.

    """

    n_samples = len(y_true)

    retval = np.stack(y_pred, axis=1)

    return retval[:n_samples]

    return y_pred[:n_samples, :, 1]

print("Evaluating model")

train_predictions = tg.predict_on_generator(data_generator(train_dataset, predict=True))

train_predictions = model.predict_on_generator(data_generator(train_dataset, predict=True))

train_predictions = reshape_y_pred(train_dataset.y, train_predictions)

train_scores = metric.compute_metric(train_dataset.y, train_predictions, train_dataset.w)

print("Training ROC-AUC Score: %f" % train_scores)

valid_predictions = tg.predict_on_generator(data_generator(valid_dataset, predict=True))

valid_predictions = model.predict_on_generator(data_generator(valid_dataset, predict=True))

valid_predictions = reshape_y_pred(valid_dataset.y, valid_predictions)

valid_scores = metric.compute_metric(valid_dataset.y, valid_predictions, valid_dataset.w)

print("Valid ROC-AUC Score: %f" % valid_scores)

```

%% Output

    Evaluating model

    computed_metrics: [0.83463194036351052, 0.86218739964675661, 0.84894031662657832, 0.80217986671584707, 0.70559942152332189, 0.79751934844253025, 0.8103057689046107, 0.71659210162938414, 0.80849247997327445, 0.72071717294380933, 0.83433314746710274, 0.78304357554399506]

    Training ROC-AUC Score: 0.793712

    computed_metrics: [0.78221936377578793, 0.78993055555555547, 0.81705388431256543, 0.77777071682765631, 0.66802272727272727, 0.67197702777122181, 0.64295604015230179, 0.72305596655628368, 0.74692724275959499, 0.63050611290902547, 0.80023473616134511, 0.73880275624461667]

    Valid ROC-AUC Score: 0.732455

    computed_metrics: [0.7500439569332503]

    Training ROC-AUC Score: 0.750044

    computed_metrics: [0.6853516799391368]

    Valid ROC-AUC Score: 0.685352

%% Cell type:markdown id: tags:

Success! The model we've constructed behaves nearly identically to `GraphConvModel`. If you're looking to build your own custom models, you can follow the example we've provided here to do so. We hope to see exciting constructions from your end soon!

%% Cell type:code id: tags:

``` python

```

    "for i in range(num_epochs):\n",

    " loss = model.fit(train_dataset, nb_epoch=1)\n",

    " print(\"Epoch %d loss: %f\" % (i, loss))\n",

    " losses.append(loss)\n",

    "model.save()"

    " losses.append(loss)"

   ]

  },

  {

   "name": "python",

   "nbconvert_exporter": "python",

   "pygments_lexer": "ipython3",

   "version": "3.5.5"

   "version": "3.6.9"

  }

 },

 "nbformat": 4,