Merge pull request #145 from openvax/v1.3

pan allele prediction (MHCflurry 1.3.0)

mhcflurry/percent_rank_transform.py

+"""
+Class for transforming arbitrary values into percent ranks given a distribution.
+"""
 import numpy
 import pandas
 
+
 class PercentRankTransform(object):
     """
     Transform arbitrary values into percent ranks.
@@ -76,8 +80,3 @@ class PercentRankTransform(object):
         result.cdf = series.values
         result.bin_edges = series.index.values[1:-1]
         return result
-
-
-
-
-

mhcflurry/predict_command.py

@@ -149,6 +149,10 @@ implementation_args.add_argument(
 
 
 def run(argv=sys.argv[1:]):
+    if not argv:
+        parser.print_help()
+        parser.exit(1)
+
     args = parser.parse_args(argv)
 
     set_keras_backend(backend=args.backend, num_threads=args.threads)
@@ -159,9 +163,9 @@ def run(argv=sys.argv[1:]):
 
     models_dir = args.models
     if models_dir is None:
-        # The reason we set the default here instead of in the argument parser is that
-        # we want to test_exists at this point, so the user gets a message instructing
-        # them to download the models if needed.
+        # The reason we set the default here instead of in the argument parser
+        # is that we want to test_exists at this point, so the user gets a
+        # message instructing them to download the models if needed.
         models_dir = get_default_class1_models_dir(test_exists=True)
     predictor = Class1AffinityPredictor.load(models_dir)
 
@@ -215,12 +219,13 @@ def run(argv=sys.argv[1:]):
         })
         logging.info(
             "Predicting for %d alleles and %d peptides = %d predictions" % (
-            len(args.alleles), len(args.peptides), len(df)))
+                len(args.alleles), len(args.peptides), len(df)))
 
     predictions = predictor.predict_to_dataframe(
         peptides=df[args.peptide_column].values,
         alleles=df[args.allele_column].values,
-        include_individual_model_predictions=args.include_individual_model_predictions,
+        include_individual_model_predictions=(
+            args.include_individual_model_predictions),
         throw=not args.no_throw)
 
     for col in predictions.columns:

mhcflurry/regression_target.py

@@ -14,7 +14,7 @@ def from_ic50(ic50, max_ic50=50000.0):
     numpy.array of float
 
     """
-    x = 1.0 - (numpy.log(ic50) / numpy.log(max_ic50))
+    x = 1.0 - (numpy.log(numpy.maximum(ic50, 1e-12)) / numpy.log(max_ic50))
     return numpy.minimum(
         1.0,
         numpy.maximum(0.0, x))

mhcflurry/scoring.py

+"""
+Measures of prediction accuracy
+"""
 from __future__ import (
     print_function,
     division,

mhcflurry/select_allele_specific_models_command.py

@@ -22,7 +22,7 @@ tqdm.monitor_interval = 0  # see https://github.com/tqdm/tqdm/issues/481
 from .class1_affinity_predictor import Class1AffinityPredictor
 from .encodable_sequences import EncodableSequences
 from .common import configure_logging, random_peptides
-from .parallelism import worker_pool_with_gpu_assignments_from_args, add_worker_pool_args
+from .local_parallelism import worker_pool_with_gpu_assignments_from_args, add_local_parallelism_args
 from .regression_target import from_ic50
 
 
@@ -176,7 +176,7 @@ parser.add_argument(
     help="Keras verbosity. Default: %(default)s",
     default=0)
 
-add_worker_pool_args(parser)
+add_local_parallelism_args(parser)
 
 
 def run(argv=sys.argv[1:]):
@@ -205,7 +205,7 @@ def run(argv=sys.argv[1:]):
         df = pandas.read_csv(args.data)
         print("Loaded data: %s" % (str(df.shape)))
 
-        df = df.ix[
+        df = df.loc[
             (df.peptide.str.len() >= 8) & (df.peptide.str.len() <= 15)
         ]
         print("Subselected to 8-15mers: %s" % (str(df.shape)))

mhcflurry/testing_utils.py

+"""
+Utilities used in MHCflurry unit tests.
+"""
+from . import Class1NeuralNetwork
+from .common import set_keras_backend
+
+
+def startup():
+    """
+    Configure Keras backend for running unit tests.
+    """
+    set_keras_backend("tensorflow-cpu", num_threads=2)
+
+
+def cleanup():
+    """
+    Clear tensorflow session and other process-wide resources.
+    """
+    import keras.backend as K
+    Class1NeuralNetwork.clear_model_cache()
+    K.clear_session()

mhcflurry/version.py

-__version__ = "1.2.4"
+__version__ = "1.3.0"

requirements.txt

@@ -8,3 +8,4 @@ scikit-learn
 mhcnames
 pyyaml
 tqdm
+np_utils
\ No newline at end of file

setup.py

@@ -58,11 +58,8 @@ if __name__ == '__main__':
         'mhcnames',
         'pyyaml',
         'tqdm',
+        'np_utils',
     ]
-    if PY2:
-        # concurrent.futures is a standard library in Py3 but Py2
-        # requires this backport
-        required_packages.append('futures')
 
     setup(
         name='mhcflurry',
@@ -78,14 +75,20 @@ if __name__ == '__main__':
                 'mhcflurry-predict = mhcflurry.predict_command:run',
                 'mhcflurry-class1-train-allele-specific-models = '
                     'mhcflurry.train_allele_specific_models_command:run',
+                'mhcflurry-class1-train-pan-allele-models = '
+                    'mhcflurry.train_pan_allele_models_command:run',
                 'mhcflurry-class1-select-allele-specific-models = '
                     'mhcflurry.select_allele_specific_models_command:run',
+                'mhcflurry-class1-select-pan-allele-models = '
+                    'mhcflurry.select_pan_allele_models_command:run',
                 'mhcflurry-calibrate-percentile-ranks = '
                     'mhcflurry.calibrate_percentile_ranks_command:run',
+                '_mhcflurry-cluster-worker-entry-point = '
+                    'mhcflurry.cluster_parallelism:worker_entry_point',
             ]
         },
         classifiers=[
-            'Development Status :: 4 - Beta',
+            'Development Status :: 5 - Production/Stable',
             'Environment :: Console',
             'Operating System :: OS Independent',
             'Intended Audience :: Science/Research',