def setup():
global ALLELE_SPECIFIC_PREDICTOR, PAN_ALLELE_PREDICTOR
startup()
ALLELE_SPECIFIC_PREDICTOR = Class1AffinityPredictor.load(
get_path("models_class1", "models"))
PAN_ALLELE_PREDICTOR = Class1AffinityPredictor.load(
get_path("models_class1_pan", "models.with_mass_spec"))
def setup():
global PREDICTORS
startup()
PREDICTORS = {
'allele-specific':
Class1AffinityPredictor.load(get_path("models_class1", "models")),
'pan-allele':
Class1AffinityPredictor.load(
get_path("models_class1_pan", "models.combined"))
}
def setup():
global PAN_ALLELE_PREDICTOR
startup()
PAN_ALLELE_PREDICTOR = Class1AffinityPredictor.load(
get_path("models_class1_pan", "models.combined"),
optimization_level=0,
)
def run_and_check(n_jobs=0, delete=True, additional_args=[]):
source_models_dir = get_path("models_class1_pan", "models.combined")
dest_models_dir = tempfile.mkdtemp(prefix="mhcflurry-test-models")
# Save a new predictor that has no percent rank calibration data.
original_predictor = Class1AffinityPredictor.load(source_models_dir)
print("Loaded predictor", source_models_dir)
new_predictor = Class1AffinityPredictor(
class1_pan_allele_models=original_predictor.class1_pan_allele_models,
allele_to_sequence=original_predictor.allele_to_sequence,
)
new_predictor.save(dest_models_dir)
print("Saved predictor to", dest_models_dir)
new_predictor = Class1AffinityPredictor.load(dest_models_dir)
assert_equal(len(new_predictor.allele_to_percent_rank_transform), 0)
args = [
"mhcflurry-calibrate-percentile-ranks",
"--models-dir",
dest_models_dir,
"--match-amino-acid-distribution-data",
get_path("data_curated", "curated_training_data.affinity.csv.bz2"),
"--motif-summary",
"--num-peptides-per-length",
"1000",
"--allele",
"HLA-A*02:01",
"HLA-B*07:02",
"--verbosity",
"1",
"--num-jobs",
str(n_jobs),
] + additional_args
print("Running with args: %s" % args)
subprocess.check_call(args)
new_predictor = Class1AffinityPredictor.load(dest_models_dir)
assert_equal(len(new_predictor.allele_to_percent_rank_transform), 2)
if delete:
print("Deleting: %s" % dest_models_dir)
shutil.rmtree(dest_models_dir)
else:
print("Not deleting: %s" % dest_models_dir)
def initialise_prediction():
"""
Enable mhcflurry prediction
:return:
"""
os.environ[
'TF_CPP_MIN_LOG_LEVEL'] = '2' # Prevents warnings. No, I'm not building from source
from mhcflurry import Class1AffinityPredictor
global predictor
predictor = Class1AffinityPredictor.load()
def __init__(self,
alleles,
default_peptide_lengths=[9],
predictor=None,
models_path=None):
"""
Parameters
-----------
alleles : list of str
default_peptide_lengths : list of int
predictor : mhcflurry.Class1AffinityPredictor (optional)
MHCflurry predictor to use
models_path : string
Models dir to use if predictor argument is None
"""
# moving import here since the mhcflurry package imports
# Keras and its backend (either Theano or TF) which end up
# slowing down responsive for any CLI application using MHCtools
from mhcflurry import Class1AffinityPredictor
BasePredictor.__init__(self,
alleles=alleles,
default_peptide_lengths=default_peptide_lengths,
min_peptide_length=8,
max_peptide_length=15)
if predictor:
self.predictor = predictor
elif models_path:
logging.info("Loading MHCflurry models from %s" % models_path)
self.predictor = Class1AffinityPredictor.load(models_path)
else:
self.predictor = Class1AffinityPredictor.load()
# relying on BasePredictor and MHCflurry to both normalize
# allele names the same way using mhcnames
for allele in self.alleles:
if allele not in self.predictor.supported_alleles:
raise UnsupportedAllele(allele)
def predict(self, peptides, alleles=None, binary=False, **kwargs):
# test whether one peptide or a list
if not isinstance(peptides, list):
peptides = [peptides]
# if no alleles are specified do predictions for all supported alleles
if alleles is None:
alleles = self.supportedAlleles
else:
# filter for supported alleles
alleles = filter(lambda a: a in self.supportedAlleles, alleles)
alleles = self.convert_alleles(alleles)
# test mhcflurry models are available => download if not
p = subprocess.Popen(
['mhcflurry-downloads', 'path', 'models_class1'],
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
if p is not 0:
subprocess.call(
['mhcflurry-downloads', 'fetch', 'models_class1'])
# load model
predictor = Class1AffinityPredictor.load()
# predict and assign binding affinities
result = {}
for a in alleles:
allele_repr = self.revert_allele_repr(a)
result[allele_repr] = {}
for p in peptides:
seq = p.__str__()
binding_affinity = predictor.predict(allele=a,
peptides=[seq])[0]
if binary:
if binding_affinity <= 500:
result[allele_repr][p] = 1.0
else:
result[allele_repr][p] = 0.0
else:
result[allele_repr][p] = binding_affinity
# create EpitopePredictionResult object. This is a multi-indexed DataFrame
# with Peptide and Method as multi-index and alleles as columns
df_result = EpitopePredictionResult.from_dict(result)
df_result.index = pandas.MultiIndex.from_tuples(
[tuple((i, self.name)) for i in df_result.index],
names=['Seq', 'Method'])
return df_result
def predict(self, sequence=None, peptides=None, length=11, overlap=1,
allele='HLA-A0101', name='', **kwargs):
"""Uses mhcflurry python classes for prediction"""
self.sequence = sequence
from mhcflurry import Class1AffinityPredictor
predictor = Class1AffinityPredictor.load()
if peptides == None:
peptides, s = peptutils.create_fragments(seq=sequence,
length=length, overlap=overlap)
df = predictor.predict_to_dataframe(peptides=peptides, allele=allele)
#print (df[:5])
df = self.prepareData(df, name, allele)
self.data = df
return df
def run_and_check(n_jobs=0):
models_dir = tempfile.mkdtemp(prefix="mhcflurry-test-models")
hyperparameters_filename = os.path.join(models_dir, "hyperparameters.yaml")
with open(hyperparameters_filename, "w") as fd:
json.dump(HYPERPARAMETERS, fd)
args = [
"mhcflurry-class1-train-allele-specific-models",
"--data",
get_path("data_curated", "curated_training_data.affinity.csv.bz2"),
"--hyperparameters",
hyperparameters_filename,
"--allele",
"HLA-A*02:01",
"HLA-A*03:01",
"--out-models-dir",
models_dir,
"--num-jobs",
str(n_jobs),
]
print("Running with args: %s" % args)
subprocess.check_call(args)
# Calibrate percentile ranks
args = [
"mhcflurry-calibrate-percentile-ranks",
"--models-dir",
models_dir,
"--num-peptides-per-length",
"10000",
"--num-jobs",
str(n_jobs),
]
print("Running with args: %s" % args)
subprocess.check_call(args)
result = Class1AffinityPredictor.load(models_dir)
predictions = result.predict(peptides=["SLYNTVATL"],
alleles=["HLA-A*02:01"])
assert_equal(predictions.shape, (1, ))
assert_array_less(predictions, 1000)
df = result.predict_to_dataframe(peptides=["SLYNTVATL"],
alleles=["HLA-A*02:01"])
print(df)
assert "prediction_percentile" in df.columns
print("Deleting: %s" % models_dir)
shutil.rmtree(models_dir)
def test_mhcflurry():
predictor = MHCflurry(alleles=[DEFAULT_ALLELE])
binding_predictions = predictor.predict_subsequences(protein_sequence_dict,
peptide_lengths=[9])
eq_(4, len(binding_predictions),
"Expected 4 binding predictions from %s" % (binding_predictions, ))
prediction_scores = {(x.peptide, x.allele): x.affinity
for x in binding_predictions}
predictor = Class1AffinityPredictor.load()
# test one prediction at a time to make sure there's no peptide/allele mixup
for (peptide, allele), affinity in prediction_scores.items():
prediction = predictor.predict([peptide], allele=allele)
assert len(prediction) == 1
# we've seen results differ a bit so doing an approximate check, not an error condition
testing.assert_almost_equal(prediction[0], affinity, decimal=0)
def setup():
global AFFINITY_PREDICTOR
global CLEAVAGE_PREDICTOR
global CLEAVAGE_PREDICTOR_NO_FLANKING
global PRESENTATION_PREDICTOR
startup()
AFFINITY_PREDICTOR = Class1AffinityPredictor.load(get_path(
"models_class1_pan", "models.combined"),
optimization_level=0,
max_models=1)
CLEAVAGE_PREDICTOR = Class1ProcessingPredictor.load(get_path(
"models_class1_processing", "models.selected.with_flanks"),
max_models=1)
CLEAVAGE_PREDICTOR_NO_FLANKING = Class1ProcessingPredictor.load(
get_path("models_class1_processing", "models.selected.no_flank"),
max_models=1)
PRESENTATION_PREDICTOR = Class1PresentationPredictor.load()
def do_predictions_mhcflurry(work_item_dicts, constant_data=None):
"""
Each dict of work items should have keys: work_item_num, peptides, alleles
"""
# This may run on the cluster in a way that misses all top level imports,
# so we have to re-import everything here.
import time
from mhcflurry.encodable_sequences import EncodableSequences
from mhcflurry import Class1AffinityPredictor
if constant_data is None:
constant_data = GLOBAL_DATA
args = constant_data['args']
assert args.predictor == "mhcflurry"
assert constant_data['cols'] == ["affinity"]
predictor = Class1AffinityPredictor.load(args.mhcflurry_models_dir)
results = []
for (i, d) in enumerate(work_item_dicts):
work_item_num = d['work_item_num']
peptides = d['peptides']
alleles = d['alleles']
print("Processing work item", i + 1, "of", len(work_item_dicts))
result = {}
results.append((work_item_num, result))
start = time.time()
peptides = EncodableSequences.create(peptides)
for (i, allele) in enumerate(alleles):
print("Processing allele %d / %d: %0.2f sec elapsed" %
(i + 1, len(alleles), time.time() - start))
for col in ["affinity"]:
result["%s %s" % (allele, col)] = predictor.predict(
peptides=peptides,
allele=allele,
throw=False,
model_kwargs={
'batch_size': args.mhcflurry_batch_size,
}).astype(constant_data['args'].result_dtype)
print("Done predicting in", time.time() - start, "sec")
return results
def predict(self, input_file, allele, epitope_length, iedb_executable_path,
iedb_retries):
predictor = Class1AffinityPredictor.load()
results = pd.DataFrame()
for line in input_file:
match = re.search('^>([0-9]+)$', line)
if match:
seq_num = match.group(1)
else:
epitopes = self.determine_neoepitopes(line.rstrip(),
epitope_length)
df = predictor.predict_to_dataframe(allele=allele,
peptides=epitopes)
df['seq_num'] = seq_num
df['start'] = df.index + 1
df.rename(columns={
'prediction': 'ic50',
'prediction_percentile': 'percentile'
},
inplace=True)
results = results.append(df)
return (results, 'pandas')
seqs = [
l.split()[1] for l in mztab_read if l.startswith("PSM")
if l.split()[1] not in seqs_new_smaller_qval
]
seqs_new_greater_qval = list(set(seqs))
seqs_new_greater_qval = [
s for s in seqs_new_greater_qval if 7 < len(s) < 13 and not 'U' in s
and not 'X' in s and not 'Z' in s and not 'J' in s and not 'B' in s
]
#call mhcflurry
#subprocess.call("mhcflurry-predict --peptides {p} --alleles {a} --out {o}".format(p=" ".join(seqs_new), a=" ".join(alleles), o=sys.argv[-1]))
seqs_filtered = []
for allele in alleles:
print(allele)
predictor = Class1AffinityPredictor.load()
df_pred = predictor.predict_to_dataframe(allele=allele,
peptides=seqs_new_greater_qval)
seqs_filtered += df_pred[df_pred['prediction'] <= float(
sys.argv[-5])]['peptide'].values.tolist()
#merge sequence lists and append decoys
seqs_new_all = list(set(seqs_new_smaller_qval + seqs_filtered))
seqs_new_all = seqs_new_all + [s[::-1] for s in seqs_new_all]
#write idXML for filtering
op = open(sys.argv[-1], 'w')
op.write(
'<PeptideIdentification score_type="q-value" higher_score_better="false">'
+ '\n')
for pep in seqs_new_all:
def test_a1_known_epitopes_in_newly_trained_model():
allele = "HLA-A*01:01"
df = pandas.read_csv(
get_path(
"data_curated", "curated_training_data.affinity.csv.bz2"))
df = df.loc[
(df.allele == allele) &
(df.peptide.str.len() >= 8) &
(df.peptide.str.len() <= 15)
]
hyperparameters = {
"max_epochs": 100,
"patience": 10,
"early_stopping": True,
"validation_split": 0.2,
"random_negative_rate": 0.0,
"random_negative_constant": 25,
"peptide_amino_acid_encoding": "BLOSUM62",
"use_embedding": False,
"kmer_size": 15,
"batch_normalization": False,
"locally_connected_layers": [
{
"filters": 8,
"activation": "tanh",
"kernel_size": 3
}
],
"activation": "relu",
"output_activation": "sigmoid",
"layer_sizes": [
32
],
"random_negative_affinity_min": 20000.0,
"random_negative_affinity_max": 50000.0,
"dense_layer_l1_regularization": 0.001,
"dropout_probability": 0.0
}
predictor = Class1AffinityPredictor()
predictor.fit_allele_specific_predictors(
n_models=2,
architecture_hyperparameters_list=[hyperparameters],
allele=allele,
peptides=df.peptide.values,
affinities=df.measurement_value.values,
verbose=0,
)
predict_and_check("HLA-A*01:01", "EVDPIGHLY", predictor=predictor)
models_dir = tempfile.mkdtemp("_models")
print(models_dir)
predictor.save(models_dir)
predictor2 = Class1AffinityPredictor.load(models_dir)
predict_and_check("HLA-A*01:01", "EVDPIGHLY", predictor=predictor2)
shutil.rmtree(models_dir)
predictor3 = Class1AffinityPredictor(
allele_to_allele_specific_models={
allele: [predictor.allele_to_allele_specific_models[allele][0]]
})
predict_and_check("HLA-A*01:01", "EVDPIGHLY", predictor=predictor3)
models_dir = tempfile.mkdtemp("_models")
print(models_dir)
predictor3.save(models_dir)
predictor4 = Class1AffinityPredictor.load(models_dir)
predict_and_check("HLA-A*01:01", "EVDPIGHLY", predictor=predictor4)
shutil.rmtree(models_dir)
def setup():
global DOWNLOADED_PREDICTOR
startup()
DOWNLOADED_PREDICTOR = Class1AffinityPredictor.load()
logging.basicConfig(level=logging.DEBUG)
import traceback
import sys
import numpy
import pandas
numpy.random.seed(0)
from mhcflurry import Class1AffinityPredictor
from nose.tools import eq_, assert_raises
from numpy import testing
from mhcflurry.downloads import get_path
from mhcflurry.testing_utils import cleanup, startup
DOWNLOADED_PREDICTOR = Class1AffinityPredictor.load()
def setup():
global DOWNLOADED_PREDICTOR
startup()
DOWNLOADED_PREDICTOR = Class1AffinityPredictor.load()
logging.basicConfig(level=logging.DEBUG)
def teardown():
global DOWNLOADED_PREDICTOR
DOWNLOADED_PREDICTOR = None
cleanup()
def check_allele_valid(self, allele):
valid_alleles = self.valid_allele_names()
if allele not in valid_alleles:
sys.exit(
"Allele %s not valid for method %s. Run `pvacseq valid_alleles %s` for a list of valid allele names."
% (allele, self.__class__.__name__, self.__class__.__name__))
class MHCI(PredictionClass, metaclass=ABCMeta):
@property
def needs_epitope_length(self):
return True
mhcflurry_predictor = Class1AffinityPredictor.load()
class MHCflurry(MHCI):
def valid_allele_names(self):
return mhcflurry_predictor.supported_alleles
def check_length_valid_for_allele(self, length, allele):
return True
def valid_lengths_for_allele(self, allele):
return [8, 9, 10, 11, 12, 13, 14]
def determine_neoepitopes(self, sequence, length):
epitopes = []
for i in range(0, len(sequence) - length + 1):
def run_and_check(n_jobs=0, delete=True, additional_args=[]):
models_dir = tempfile.mkdtemp(prefix="mhcflurry-test-models")
hyperparameters_filename = os.path.join(models_dir, "hyperparameters.yaml")
with open(hyperparameters_filename, "w") as fd:
json.dump(HYPERPARAMETERS_LIST, fd)
data_df = pandas.read_csv(
get_path("data_curated", "curated_training_data.no_mass_spec.csv.bz2"))
selected_data_df = data_df.loc[data_df.allele.str.startswith("HLA-A")]
selected_data_df.to_csv(os.path.join(models_dir, "_train_data.csv"),
index=False)
args = [
"mhcflurry-class1-train-pan-allele-models",
"--data",
os.path.join(models_dir, "_train_data.csv"),
"--allele-sequences",
get_path("allele_sequences", "allele_sequences.csv"),
"--hyperparameters",
hyperparameters_filename,
"--out-models-dir",
models_dir,
"--num-jobs",
str(n_jobs),
"--num-folds",
"2",
"--verbosity",
"1",
] + additional_args
print("Running with args: %s" % args)
subprocess.check_call(args)
# Run model selection
models_dir_selected = tempfile.mkdtemp(
prefix="mhcflurry-test-models-selected")
args = [
"mhcflurry-class1-select-pan-allele-models",
"--data",
os.path.join(models_dir, "train_data.csv.bz2"),
"--models-dir",
models_dir,
"--out-models-dir",
models_dir_selected,
"--max-models",
"1",
"--num-jobs",
str(n_jobs),
] + additional_args
print("Running with args: %s" % args)
subprocess.check_call(args)
result = Class1AffinityPredictor.load(models_dir_selected,
optimization_level=0)
assert_equal(len(result.neural_networks), 2)
predictions = result.predict(peptides=["SLYNTVATL"],
alleles=["HLA-A*02:01"])
assert_equal(predictions.shape, (1, ))
assert_array_less(predictions, 1000)
if delete:
print("Deleting: %s" % models_dir)
shutil.rmtree(models_dir)
shutil.rmtree(models_dir_selected)
def valid_allele_names(self):
predictor = Class1AffinityPredictor.load()
return predictor.supported_alleles
def setup():
global DOWNLOADED_PREDICTOR
startup()
DOWNLOADED_PREDICTOR = Class1AffinityPredictor.load()
def predict(self, peptides, alleles=None, binary=False, **kwargs):
# test whether one peptide or a list
if not isinstance(peptides, list):
peptides = [peptides]
# if no alleles are specified do predictions for all supported alleles
if alleles is None:
alleles = self.supportedAlleles
else:
# filter for supported alleles
alleles = filter(lambda a: a in self.supportedAlleles, alleles)
alleles = self.convert_alleles(alleles)
# test mhcflurry models are available => download if not
p = subprocess.Popen(
['mhcflurry-downloads', 'path', 'models_class1'],
stdout=subprocess.PIPE,
stderr=subprocess.PIPE)
if p is not 0:
subprocess.call(
['mhcflurry-downloads', 'fetch', 'models_class1'])
# load model
predictor = Class1AffinityPredictor.load()
# prepare results dictionary
result = defaultdict(defaultdict)
# group peptides by length
peptides.sort(key=len)
for length, peps in itertools.groupby(peptides, key=len):
if length not in self.supportedLength:
logging.warn(
"Peptide length must be at least %i or at most %i for %s but is %i"
% (min(self.supportedLength), max(
self.supportedLength), self.name, length))
continue
peps = list(peps)
# predict and assign binding affinities
for a in alleles:
allele_repr = self.revert_allele_repr(a)
for p in peps:
binding_affinity = predictor.predict(allele=a,
peptides=[str(p)
])[0]
if binary:
if binding_affinity <= 500:
result[allele_repr][p] = 1.0
else:
result[allele_repr][p] = 0.0
else:
result[allele_repr][p] = binding_affinity
if not result:
raise ValueError(
"No predictions could be made with " + self.name +
" for given input. Check your epitope length and HLA allele combination."
)
# create EpitopePredictionResult object. This is a multi-indexed DataFrame
# with Peptide and Method as multi-index and alleles as columns
df_result = EpitopePredictionResult.from_dict(result)
df_result.index = pandas.MultiIndex.from_tuples(
[tuple((i, self.name)) for i in df_result.index],
names=['Seq', 'Method'])
return df_result
import numpy as np
import pandas as pd
from pathlib import Path
from mhcflurry import Class1AffinityPredictor
from tqdm import tqdm
import seaborn as sns
import matplotlib.pyplot as plt
# Load data and models - change path as necessary
DATA = Path("./data")
ms_file = DATA / "abelin_peptides.mhcflurry_no_mass_spec.csv"
MODELS = Path("/Users/haemin/Library/Application Support/mhcflurry/4/1.4.0")
no_ms_model = MODELS / "models_class1_selected_no_mass_spec/models"
no_ms_predictor = Class1AffinityPredictor.load(no_ms_model)
# Read data file and initial cleanup
ms = pd.read_csv(ms_file)
ms = ms.rename(columns={"mhcflurry": "mhcflurry2"})
ms["mhcflurry4"] = np.full_like(ms.mhcflurry2, -1)
ms = ms.loc[ms.allele.isin(no_ms_predictor.supported_alleles), :]
ms["peptide_len"] = ms.peptide.str.len()
ms.loc[ms["peptide_len"] > 12, "peptide_len"] = 13
'''
Generate Figure A
'''
# Compute PPV values
models = ["netmhc", "netmhcpan", "mhcflurry2", "mhcflurry4"]
alleles = ms.allele[ms.allele.isin(no_ms_predictor.supported_alleles)].unique()
# Compute predictions for mhcflurry4
def run_and_check_with_model_selection(n_jobs=1):
models_dir1 = tempfile.mkdtemp(prefix="mhcflurry-test-models")
hyperparameters_filename = os.path.join(models_dir1,
"hyperparameters.yaml")
# Include one architecture that has max_epochs = 0. We check that it never
# gets selected in model selection.
hyperparameters = [
deepcopy(HYPERPARAMETERS[0]),
deepcopy(HYPERPARAMETERS[0]),
]
hyperparameters[-1]["max_epochs"] = 0
with open(hyperparameters_filename, "w") as fd:
json.dump(hyperparameters, fd)
args = [
"mhcflurry-class1-train-allele-specific-models",
"--data",
get_path("data_curated", "curated_training_data.affinity.csv.bz2"),
"--hyperparameters",
hyperparameters_filename,
"--allele",
"HLA-A*02:01",
"HLA-A*03:01",
"--out-models-dir",
models_dir1,
"--num-jobs",
str(n_jobs),
"--held-out-fraction-reciprocal",
"10",
"--n-models",
"1",
]
print("Running with args: %s" % args)
subprocess.check_call(args)
result = Class1AffinityPredictor.load(models_dir1)
assert_equal(len(result.neural_networks), 4)
models_dir2 = tempfile.mkdtemp(prefix="mhcflurry-test-models")
args = [
"mhcflurry-class1-select-allele-specific-models",
"--data",
get_path("data_curated", "curated_training_data.affinity.csv.bz2"),
"--exclude-data",
models_dir1 + "/train_data.csv.bz2",
"--out-models-dir",
models_dir2,
"--models-dir",
models_dir1,
"--num-jobs",
str(n_jobs),
"--mse-max-models",
"1",
"--unselected-accuracy-scorer",
"combined:mass-spec,mse",
"--unselected-accuracy-percentile-threshold",
"95",
]
print("Running with args: %s" % args)
subprocess.check_call(args)
result = Class1AffinityPredictor.load(models_dir2)
assert_equal(len(result.neural_networks), 2)
assert_equal(len(result.allele_to_allele_specific_models["HLA-A*02:01"]),
1)
assert_equal(len(result.allele_to_allele_specific_models["HLA-A*03:01"]),
1)
assert_equal(
result.allele_to_allele_specific_models["HLA-A*02:01"]
[0].hyperparameters["max_epochs"], 500)
assert_equal(
result.allele_to_allele_specific_models["HLA-A*03:01"]
[0].hyperparameters["max_epochs"], 500)
print("Deleting: %s" % models_dir1)
print("Deleting: %s" % models_dir2)
shutil.rmtree(models_dir1)
