From df5ac4846d6ff811d3e514e9c63f129409fb7a2f Mon Sep 17 00:00:00 2001
From: Alex Rubinsteyn <alex.rubinsteyn@gmail.com>
Date: Fri, 29 Apr 2016 12:17:22 -0400
Subject: [PATCH] fixed neural network test, fixed length filtering of data
loading, started on adding imputation to training script
---
mhcflurry/data.py | 17 ++++----
mhcflurry/imputation.py | 40 ++++++++++++++++++-
...rry-train-class1-allele-specific-models.py | 26 +++++++++---
test/test_neural_nets.py | 17 +++++---
4 files changed, 80 insertions(+), 20 deletions(-)
diff --git a/mhcflurry/data.py b/mhcflurry/data.py
index eab00b13..1ea71212 100644
--- a/mhcflurry/data.py
+++ b/mhcflurry/data.py
@@ -68,13 +68,12 @@ def _infer_csv_separator(filename):
def load_dataframe(
filename,
- peptide_length=None,
max_ic50=MAX_IC50,
sep=None,
species_column_name="species",
allele_column_name="mhc",
peptide_column_name=None,
- peptide_length_column_name="peptide_length",
+ filter_peptide_length=None,
ic50_column_name="meas",
only_human=False):
"""
@@ -88,9 +87,6 @@ def load_dataframe(
- 'sequence'
- 'meas'
- peptide_length : int, optional
- Which length peptides to use (default=load all lengths)
-
max_ic50 : float
Treat IC50 scores above this value as all equally bad
(transform them to 0.0 in the regression output)
@@ -101,6 +97,9 @@ def load_dataframe(
peptide_column_name : str, optional
Default behavior is to try {"sequence", "peptide", "peptide_sequence"}
+ filter_peptide_length : int, optional
+ Which length peptides to use (default=load all lengths)
+
only_human : bool
Only load entries from human MHC alleles
@@ -131,8 +130,9 @@ def load_dataframe(
if only_human:
human_mask = df[species_column_name] == "human"
df = df[human_mask]
- if peptide_length is not None:
- length_mask = df[peptide_length_column_name] == peptide_length
+
+ if filter_peptide_length:
+ length_mask = df[peptide_column_name].str.len() == filter_peptide_length
df = df[length_mask]
df[allele_column_name] = df[allele_column_name].map(normalize_allele_name)
@@ -383,11 +383,10 @@ def load_allele_datasets(
filename=filename,
max_ic50=max_ic50,
sep=sep,
- peptide_length=peptide_length,
species_column_name=species_column_name,
allele_column_name=allele_column_name,
peptide_column_name=peptide_column_name,
- peptide_length_column_name=peptide_length_column_name,
+ filter_peptide_length=None if use_multiple_peptide_lengths else peptide_length,
ic50_column_name=ic50_column_name,
only_human=only_human)
diff --git a/mhcflurry/imputation.py b/mhcflurry/imputation.py
index 145c6e9a..1bcb08dc 100644
--- a/mhcflurry/imputation.py
+++ b/mhcflurry/imputation.py
@@ -19,7 +19,16 @@ from __future__ import (
)
from collections import defaultdict
import numpy as np
-from fancyimpute.dictionary_helpers import dense_matrix_from_nested_dictionary
+from fancyimpute.dictionary_helpers import (
+ dense_matrix_from_nested_dictionary
+)
+from fancyimpute import (
+ KNN,
+ IterativeSVD,
+ SimpleFill,
+ SoftImpute,
+ MICE
+)
from .data import (
create_allele_data_from_peptide_to_ic50_dict,
@@ -190,3 +199,32 @@ def create_imputed_datasets(
for (allele_name, allele_dict)
in allele_to_peptide_to_affinity_dict.items()
}
+
+def imputer_from_name(imputation_method_name, **kwargs):
+ """
+ Helper function for constructing an imputation object from a name given
+ typically from a commandline argument.
+ """
+ imputation_method_name = imputation_method_name.strip().lower()
+ if imputation_method_name == "mice":
+ kwargs["n_burn_in"] = kwargs.get("n_burn_in", 5)
+ kwargs["n_imputations"] = kwargs.get("n_imputations", 25)
+ kwargs["min_value"] = kwargs.get("min_value", 0)
+ kwargs["max_value"] = kwargs.get("max_value", 1)
+ return MICE(**kwargs)
+ elif imputation_method_name == "knn":
+ kwargs["k"] = kwargs.get("k", 3)
+ kwargs["orientation"] = kwargs.get("orientation", "columns")
+ kwargs["print_interval"] = kwargs.get("print_interval", 10)
+ return KNN(**kwargs)
+ elif imputation_method_name == "svd":
+ kwargs["rank"] = kwargs.get("rank", 10)
+ return IterativeSVD(**kwargs)
+ elif imputation_method_name == "svt" or imputation_method_name == "softimpute":
+ return SoftImpute(**kwargs)
+ elif imputation_method_name == "mean":
+ return SimpleFill("mean", **kwargs)
+ elif imputation_method_name == "none":
+ return None
+ else:
+ raise ValueError("Invalid imputation method: %s" % imputation_method_name)
diff --git a/script/mhcflurry-train-class1-allele-specific-models.py b/script/mhcflurry-train-class1-allele-specific-models.py
index dab69768..7a1475e9 100755
--- a/script/mhcflurry-train-class1-allele-specific-models.py
+++ b/script/mhcflurry-train-class1-allele-specific-models.py
@@ -53,6 +53,8 @@ from mhcflurry.paths import (
CLASS1_MODEL_DIRECTORY,
CLASS1_DATA_DIRECTORY
)
+from mhcflurry.imputation import imputer_from_name, create_imputed_datasets
+
CSV_FILENAME = "combined_human_class1_dataset.csv"
CSV_PATH = join(CLASS1_DATA_DIRECTORY, CSV_FILENAME)
@@ -92,16 +94,24 @@ parser.add_argument(
nargs="+",
type=normalize_allele_name)
+parser.add_argument(
+ "--imputation-method",
+ default=None,
+ choices=("mice", "knn", "softimpute", "svd", "mean"),
+ type=imputer_from_name,
+ help="Use the given imputation method to generate data for pre-training models")
+
# add options for neural network hyperparameters
parser = add_hyperparameter_arguments_to_parser(parser)
if __name__ == "__main__":
args = parser.parse_args()
print(args)
+
if not exists(args.output_dir):
makedirs(args.output_dir)
- allele_groups = load_allele_datasets(
+ allele_data_dict = load_allele_datasets(
filename=args.binding_data_csv,
peptide_length=9,
use_multiple_peptide_lengths=True,
@@ -111,15 +121,21 @@ if __name__ == "__main__":
# concatenate datasets from all alleles to use for pre-training of
# allele-specific predictors
- X_all = np.vstack([group.X_index for group in allele_groups.values()])
- Y_all = np.concatenate([group.Y for group in allele_groups.values()])
+ X_all = np.vstack([group.X_index for group in allele_data_dict.values()])
+ Y_all = np.concatenate([group.Y for group in allele_data_dict.values()])
print("Total Dataset size = %d" % len(Y_all))
+ if args.imputation_method is not None:
+ # TODO: use imputed data for training
+ imputed_data_dict = create_imputed_datasets(
+ allele_data_dict,
+ args.imputation_method)
+
# if user didn't specify alleles then train models for all available alleles
alleles = args.alleles
if not alleles:
- alleles = sorted(allele_groups.keys())
+ alleles = sorted(allele_data_dict.keys())
for allele_name in alleles:
allele_name = normalize_allele_name(allele_name)
@@ -127,7 +143,7 @@ if __name__ == "__main__":
print("Skipping allele %s" % (allele_name,))
continue
- allele_data = allele_groups[allele_name]
+ allele_data = allele_data_dict[allele_name]
X = allele_data.X_index
Y = allele_data.Y
diff --git a/test/test_neural_nets.py b/test/test_neural_nets.py
index 99f8f774..966efdd5 100644
--- a/test/test_neural_nets.py
+++ b/test/test_neural_nets.py
@@ -11,7 +11,7 @@ def test_make_embedding_network():
activation="tanh",
embedding_input_dim=3,
embedding_output_dim=20,
- learning_rate=0.1)
+ learning_rate=0.05)
X_negative = np.array([
[0] * 3,
@@ -33,7 +33,7 @@ def test_make_embedding_network():
])
X_index = np.vstack([X_negative, X_positive])
Y = np.array([0.0] * len(X_negative) + [1.0] * len(X_positive))
- nn.fit(X_index, Y, nb_epoch=10)
+ nn.fit(X_index, Y, nb_epoch=20)
Y_pred = nn.predict(X_index)
print(Y)
print(Y_pred)
@@ -46,7 +46,7 @@ def test_make_hotshot_network():
activation="relu",
layer_sizes=[4],
n_amino_acids=2,
- learning_rate=0.1)
+ learning_rate=0.05)
X_binary = np.array([
[True, False, True, False, True, False],
[True, False, True, False, False, True],
@@ -56,9 +56,16 @@ def test_make_hotshot_network():
[False, True, True, False, False, True],
], dtype=bool)
Y = np.array([0.0, 0.0, 0.0, 0.0, 1.0, 1.0])
- nn.fit(X_binary, Y, nb_epoch=10)
+ nn.fit(X_binary, Y, nb_epoch=20)
Y_pred = nn.predict(X_binary)
print(Y)
print(Y_pred)
for (Y_i, Y_pred_i) in zip(Y, Y_pred):
- assert abs(Y_i - Y_pred_i) <= 0.25, (Y_i, Y_pred_i)
+ if Y_i:
+ assert Y_pred_i >= 0.6, "Expected higher value than %f" % Y_pred_i
+ else:
+ assert Y_pred_i <= 0.4, "Expected lower value than %f" % Y_pred_i
+
+if __name__ == "__main__":
+ test_make_hotshot_network()
+ test_make_embedding_network()
--
GitLab