Fixing more examples

HPPB_URL = "http://deepchem.io.s3-website-us-west-1.amazonaws.com/datasets/hppb.csv"

DEFAULT_DATA_DIR = deepchem.utils.get_data_dir()

def remove_missing_entries(dataset):

  """Remove missing entries.

  Some of the datasets have missing entries that sneak in as zero'd out

  feature vectors. Get rid of them.

  """

  for i, (X, y, w, ids) in enumerate(dataset.itershards()):

    ###########################################

    print("X.shape")

    print(X.shape)

    print("X[:10]")

    print(X[:10])

    print("type(X)")

    print(type(X))

    ###########################################

    #available_rows = X.any(axis=1)

    available_rows = np.any(X)

    ###########################################

    print("type(available_rows)")

    print(type(available_rows))

    ###########################################

    logger.info("Shard %d has %d missing entries." %

                #(i, np.count_nonzero(~available_rows)))

                (i, len(X) - len(available_rows)))

    ##################################################

    print("type(available_rows)")

    print(type(available_rows))

    ##################################################

    X = X[available_rows]

    y = y[available_rows]

    w = w[available_rows]

    ids = ids[available_rows]

    dataset.set_shard(i, X, y, w, ids)

def load_hppb(featurizer="ECFP",

              data_dir=None,

              save_dir=None,

  dataset_file = os.path.join(data_dir, "hppb.csv")

  if not os.path.exists(dataset_file):

    logger.info("{} does not exist. Downloading it.".format(dataset_file))

    ################################################

    print("data_dir")

    print(data_dir)

    ################################################

    deepchem.utils.download_url(url=HPPB_URL, dest_dir=data_dir)

  if featurizer == 'ECFP':

  dataset = loader.featurize(input_files=[dataset_file], shard_size=2000)

  logger.info("Removing missing entries...")

  remove_missing_entries(dataset)

  if split == None:

    logger.info("About to transform the data...")

# ADME Dataset Examples

ADME (Absorption, Distribution, Metabolism, Excretion) is a core

part of the drug discovery process. In-silico models for ADME

tasks span a wide variety of pharmacokinetics endpoints across

multiple species.

ADME (Absorption, Distribution, Metabolism, Excretion) is a

core part of the drug discovery process. In-silico models for

ADME tasks span a wide variety of pharmacokinetics endpoints

across multiple species.

The ADME benchmark contains three of the larger datasets that

were released by AstraZeneca on ChEMBL: human plasma protein

MODEL = "GraphConv"

SPLIT = "scaffold"

DATASET_NAME = "hppb"

DATASET_NAME = "clearance"

BATCH_SIZE = 128

# Set to higher values to get better numbers

MAX_EPOCH = 1

#def retrieve_datasets():

#  os.system(

#      'wget -c %s' %

#      'https://s3-us-west-1.amazonaws.com/deep-crystal-california/az_logd.csv')

#  os.system(

#      'wget -c %s' %

#      'https://s3-us-west-1.amazonaws.com/deep-crystal-california/az_hppb.csv')

#  os.system(

#      'wget -c %s' %

#      'https://s3-us-west-1.amazonaws.com/deep-crystal-california/az_clearance.csv'

#  )

MAX_EPOCH = 10

def load_dataset(dataset_name, featurizer='ECFP', split='index'):

  #tasks = ['exp']

  #if featurizer == 'ECFP':

  #  featurizer = dc.feat.CircularFingerprint(size=1024)

  #elif featurizer == 'GraphConv':

  #  featurizer = dc.feat.ConvMolFeaturizer()

  #loader = dc.data.CSVLoader(

  #    tasks=tasks, smiles_field="smiles", featurizer=featurizer)

  #dataset = loader.featurize(dataset_file, shard_size=8192)

  if dataset_name.lower() == "clearance":

    dataset = dc.molnet.load_clearance(featurizer=featurizer, split=split)

    return dc.molnet.load_clearance(featurizer=featurizer, split=split)

  elif dataset_name.lower() == "hppb":

    dataset = dc.molnet.load_hppb(featurizer=featurizer, split=split)

  #transformers = [

  #    dc.trans.NormalizationTransformer(transform_y=True, dataset=dataset)

  #]

  #for transformer in transformers:

  #  dataset = transformer.transform(dataset)

  #splitters = {

  #    'index': dc.splits.IndexSplitter(),

  #    'random': dc.splits.RandomSplitter(),

  #    'scaffold': dc.splits.ScaffoldSplitter()

  #}

  #splitter = splitters[split]

  #train, valid, test = splitter.train_valid_test_split(dataset)

  #return tasks, (train, valid, test), transformers

    return dc.molnet.load_hppb(featurizer=featurizer, split=split)

  elif dataset_name.lower() == "lipo":

    return dc.molnet.load_lipo(featurizer=featurizer, split=split)

def experiment(dataset_name, method='GraphConv', split='scaffold'):

#======================================================================

# Run Benchmarks {GC-DNN, SVR, RF}

def main():

  #print("About to retrieve datasets")

  #retrieve_datasets()

  metric = dc.metrics.Metric(dc.metrics.pearson_r2_score, np.mean)

  print("About to build model")

# Dataset Description

This example is based on the DUD-E group; it contained 102 datasets that were

designed for the evaluation of methods to predict interactions between proteins

and small molecules (Mysinger et al., 2012)

This example is based on the DUD-E group; it contained 102

datasets that were designed for the evaluation of methods to

predict interactions between proteins and small molecules

Mysinger, Michael M., et al. "Directory of useful decoys, enhanced (DUD-E): better ligands and decoys for better benchmarking." Journal of medicinal chemistry 55.14 (2012): 6582-6594.

# Dataset overview

The Tox21 datasets were used in the recent (Tox21 Data Challenge)[https://tripod.nih.gov/tox21/challenge/]; they contain experimental data for targets relevant to drug toxicity prediction

The Tox21 datasets were used in the recent (Tox21 Data

Challenge)[https://tripod.nih.gov/tox21/challenge/]; they contain experimental

data for targets relevant to drug toxicity prediction

Ref: B Ramsundar, S Kearnes, P Riley, D Webster, D Konerding, V Pande

arXiv preprint arXiv:1502.02072

 No newline at end of file

This folder features a number of different models trained on the Tox21 dataset.