Minor fixes

 - Add default values and usage to argparse help messages
 - Require theano >=0.8.2 (was getting errors instantiating a model at 0.7.0)

README.md

@@ -19,10 +19,10 @@ Download, Normalize, and Combine Training Data:
 (make sure you have `wget` available, e.g. `brew install wget` on Mac OS X)
 
 ```
-scripts/download-iedb.sh
-scripts/download-peters-2013-dataset.sh
-scripts/create-iedb-class1-dataset.py
-scripts/create-combined-class1-dataset.py
+script/download-iedb.sh
+script/download-peters-2013-dataset.sh
+script/create-iedb-class1-dataset.py
+script/create-combined-class1-dataset.py
 ```
 
 ## Train Neural Network Models

mhcflurry/class1_allele_specific_hyperparameters.py

@@ -36,46 +36,50 @@ def add_hyperparameter_arguments_to_parser(parser):
         "--training-epochs",
         type=int,
         default=N_EPOCHS,
-        help="Number of training epochs")
+        help="Number of training epochs. Default: %(default)s")
 
     parser.add_argument(
         "--initialization",
         default=INITIALIZATION_METHOD,
-        help="Initialization for neural network weights")
+        help="Initialization for neural network weights Default: %(default)s")
 
     parser.add_argument(
         "--activation",
         default=ACTIVATION,
-        help="Activation function for neural network layers")
+        help="Activation function for neural network layers. "
+        "Default: %(default)s")
 
     parser.add_argument(
         "--embedding-size",
         type=int,
         default=EMBEDDING_DIM,
-        help="Size of vector representations for embedding amino acids")
+        help="Size of vector representations for embedding amino acids. "
+        "Default: %(default)s")
 
     parser.add_argument(
         "--hidden-layer-size",
         type=int,
         default=HIDDEN_LAYER_SIZE,
-        help="Size of hidden neural network layer")
+        help="Size of hidden neural network layer. Default: %(default)s")
 
     parser.add_argument(
         "--dropout",
         type=float,
         default=DROPOUT_PROBABILITY,
-        help="Dropout probability after neural network layers")
+        help="Dropout probability after neural network layers. "
+        "Default: %(default)s")
 
     parser.add_argument(
         "--max-ic50",
         type=float,
         default=MAX_IC50,
-        help="Largest IC50 represented by neural network output")
+        help="Largest IC50 represented by neural network output. "
+        "Default: %(default)s")
 
     parser.add_argument(
         "--learning-rate",
         type=float,
         default=0.001,
-        help="Learning rate for training neural network")
+        help="Learning rate for training neural network. Default: %(default)s")
 
     return parser

requirements.txt

 numpy>= 1.7
 pandas>=0.13.1
 appdirs
-theano
+theano>=0.8.2
 keras<1.0
 h5py
 cherrypy

script/create-combined-class1-dataset.py

@@ -41,7 +41,7 @@ IEDB_PICKLE_PATH = join(CLASS1_DATA_DIRECTORY, IEDB_PICKLE_FILENAME)
 
 KIM_2013_CSV_FILENAME = "bdata.20130222.mhci.public.1.txt"
 
-parser = argparse.ArgumentParser()
+parser = argparse.ArgumentParser(usage=__doc__)
 
 parser.add_argument(
     "--ic50-fraction-tolerance",
@@ -49,40 +49,46 @@ parser.add_argument(
     type=float,
     help=(
         "How much can the IEDB and NetMHCpan IC50 differ and still be"
-        " considered compatible (as a fraction of the NetMHCpan value)"))
+        " considered compatible (as a fraction of the NetMHCpan value). "
+        "Default: %(default)s"))
 
 parser.add_argument(
     "--min-assay-overlap-size",
     type=int,
     default=1,
-    help="Minimum number of entries overlapping between IEDB assay and NetMHCpan data")
+    help="Minimum number of entries overlapping between IEDB assay and "
+    "NetMHCpan data. Default: %(default)s")
 
 
 parser.add_argument(
     "--min-assay-fraction-same",
     type=float,
-    help="Minimum fraction of peptides whose IC50 values agree with the NetMHCpan data",
+    help="Minimum fraction of peptides whose IC50 values agree with the "
+    "NetMHCpan data. Default: %(default)s",
     default=0.9)
 
 parser.add_argument(
     "--iedb-pickle-path",
     default=IEDB_PICKLE_PATH,
-    help="Path to .pickle file containing dictionary of IEDB assay datasets")
+    help="Path to .pickle file containing dictionary of IEDB assay datasets. "
+    "Default: %(default)s")
 
 parser.add_argument(
     "--netmhcpan-csv-path",
     default=KIM_2013_CSV_FILENAME,
-    help="Path to CSV with NetMHCpan dataset from 2013 Peters paper")
+    help="Path to CSV with NetMHCpan dataset from 2013 Peters paper. "
+    "Default: %(default)s")
 
 parser.add_argument(
     "--output-dir",
     default=CLASS1_DATA_DIRECTORY,
-    help="Path to directory where output CSV should be written")
+    help="Path to directory where output CSV should be written. "
+    "Default: %(default)s")
 
 parser.add_argument(
     "--output-csv-filename",
     default=CLASS1_DATA_CSV_FILENAME,
-    help="Name of combined CSV file")
+    help="Name of combined CSV file. Default: %(default)s")
 
 parser.add_argument(
     "--extra-dataset-csv-path",

script/create-iedb-class1-dataset.py

@@ -33,25 +33,27 @@ from mhcflurry.paths import CLASS1_DATA_DIRECTORY
 IEDB_SOURCE_FILENAME = "mhc_ligand_full.csv"
 PICKLE_OUTPUT_FILENAME = "iedb_human_class1_assay_datasets.pickle"
 
-parser = argparse.ArgumentParser()
+parser = argparse.ArgumentParser(usage=__doc__)
 
 parser.add_argument(
     "--input-csv",
     default=IEDB_SOURCE_FILENAME,
-    help="CSV file with IEDB's MHC binding data")
+    help="CSV file with IEDB's MHC binding data. Default: %(default)s")
 
 parser.add_argument(
     "--output-dir",
     default=CLASS1_DATA_DIRECTORY,
-    help="Directory to write output pickle file")
+    help="Directory to write output pickle file. Default: %(default)s")
 
 parser.add_argument(
     "--output-pickle-filename",
     default=PICKLE_OUTPUT_FILENAME,
-    help="Path to .pickle file containing dictionary of IEDB assay datasets")
+    help="Path to .pickle file containing dictionary of IEDB assay datasets. "
+    "Default: %(default)s")
 
 parser.add_argument(
     "--alleles",
+    metavar="ALLELE",
     nargs="+",
     default=[],
     help="Restrict dataset to specified alleles")

script/mhcflurry-train-class1-allele-specific-models.py

@@ -61,12 +61,14 @@ parser = argparse.ArgumentParser()
 parser.add_argument(
     "--binding-data-csv",
     default=CSV_PATH,
-    help="CSV file with 'mhc', 'peptide', 'peptide_length', 'meas' columns")
+    help="CSV file with 'mhc', 'peptide', 'peptide_length', 'meas' columns. "
+    "Default: %(default)s")
 
 parser.add_argument(
     "--output-dir",
     default=CLASS1_MODEL_DIRECTORY,
-    help="Output directory for allele-specific predictor HDF weights files")
+    help="Output directory for allele-specific predictor HDF weights files. "
+    "Default: %(default)s")
 
 parser.add_argument(
     "--overwrite",
@@ -77,11 +79,15 @@ parser.add_argument(
 parser.add_argument(
     "--min-samples-per-allele",
     default=5,
-    help="Don't train predictors for alleles with fewer samples than this",
+    metavar="N",
+    help="Don't train predictors for alleles with fewer than N samples. "
+    "Default: %(default)s",
     type=int)
 
 parser.add_argument(
     "--alleles",
+    metavar="ALLELE",
+    help="Alleles to train",
     default=[],
     nargs="+",
     type=normalize_allele_name)

setup.py

@@ -64,7 +64,7 @@ if __name__ == '__main__':
             'numpy>=1.7',
             'pandas>=0.13.1',
             'appdirs',
-            'theano',
+            'theano>=0.8.2',
             'keras',
             'fancyimpute',
             'scikit-learn',