Skip to content

Commit

Permalink
Merge pull request #232 from Zethson/fix/ann_prediction
Browse files Browse the repository at this point in the history 
[FIX] ANN Predictor alleles mismatching #231
  • Loading branch information
@christopher-mohr
christopher-mohr authored Nov 20, 2019
2 parents 4d1130d + a0518f8 commit 12fcaf8
Showing 1 changed file with 85 additions and 73 deletions.
158 changes: 85 additions & 73 deletions Fred2/EpitopePrediction/ANN.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -80,24 +80,24 @@ class MHCNuggetsPredictor_1(AANNEpitopePrediction):
"""
__metaclass__ = SignatureCheckerMeta
__alleles = frozenset(
["HLA-A01:01", "HLA-A02:01", "HLA-A02:02", "HLA-A02:03", "HLA-A02:04", "HLA-A02:05", "HLA-A02:06",
"HLA-A02:07", "HLA-A02:08", "HLA-A02:09", "HLA-A02:10", "HLA-A02:11", "HLA-A02:12", "HLA-A02:14",
"HLA-A02:16", "HLA-A02:17", "HLA-A02:19", "HLA-A02:50", "HLA-A03:01", "HLA-A03:02", "HLA-A03:19",
"HLA-A11:01", "HLA-A11:02", "HLA-A23:01", "HLA-A24:01", "HLA-A24:02", "HLA-A24:03", "HLA-A25:01",
"HLA-A26:01", "HLA-A26:02", "HLA-A26:03", "HLA-A29:01", "HLA-A29:02", "HLA-A30:01", "HLA-A30:02",
"HLA-A30:03", "HLA-A30:04", "HLA-A31:01", "HLA-A32:01", "HLA-A32:07", "HLA-A32:15", "HLA-A33:01",
"HLA-A33:03", "HLA-A66:01", "HLA-A68:01", "HLA-A68:02", "HLA-A68:23", "HLA-A69:01", "HLA-A74:01",
"HLA-A80:01", "HLA-B07:01", "HLA-B07:02", "HLA-B08:01", "HLA-B08:02", "HLA-B08:03", "HLA-B12:01",
"HLA-B13:02", "HLA-B14:01", "HLA-B14:02", "HLA-B15:01", "HLA-B15:02", "HLA-B15:03", "HLA-B15:08",
"HLA-B15:09", "HLA-B15:10", "HLA-B15:13", "HLA-B15:16", "HLA-B15:17", "HLA-B15:42", "HLA-B18:01",
"HLA-B27:01", "HLA-B27:02", "HLA-B27:03", "HLA-B27:04", "HLA-B27:05", "HLA-B27:06", "HLA-B27:09",
"HLA-B27:10", "HLA-B27:20", "HLA-B35:01", "HLA-B35:02", "HLA-B35:03", "HLA-B35:08", "HLA-B37:01",
"HLA-B38:01", "HLA-B39:01", "HLA-B39:06", "HLA-B39:09", "HLA-B39:10", "HLA-B40:01", "HLA-B40:02",
"HLA-B40:13", "HLA-B41:03", "HLA-B41:04", "HLA-B42:01", "HLA-B42:02", "HLA-B44:01", "HLA-B44:02",
"HLA-B44:03", "HLA-B44:05", "HLA-B45:01", "HLA-B45:06", "HLA-B46:01", "HLA-B48:01", "HLA-B51:01",
"HLA-B51:02", "HLA-B52:01", "HLA-B53:01", "HLA-B54:01", "HLA-B55:01", "HLA-B55:02", "HLA-B56:01",
"HLA-B57:01", "HLA-B57:02", "HLA-B57:03", "HLA-B58:01", "HLA-B58:02", "HLA-B60:01", "HLA-B61:01",
"HLA-B62:01", "HLA-B73:01", "HLA-B81:01", "HLA-B83:01"])
["HLA-A*01:01", "HLA-A*02:01", "HLA-A*02:02", "HLA-A*02:03", "HLA-A*02:04", "HLA-A*02:05", "HLA-A*02:06",
"HLA-A*02:07", "HLA-A*02:08", "HLA-A*02:09", "HLA-A*02:10", "HLA-A*02:11", "HLA-A*02:12", "HLA-A*02:14",
"HLA-A*02:16", "HLA-A*02:17", "HLA-A*02:19", "HLA-A*02:50", "HLA-A*03:01", "HLA-A*03:02", "HLA-A*03:19",
"HLA-A*11:01", "HLA-A*11:02", "HLA-A*23:01", "HLA-A*24:01", "HLA-A*24:02", "HLA-A*24:03", "HLA-A*25:01",
"HLA-A*26:01", "HLA-A*26:02", "HLA-A*26:03", "HLA-A*29:01", "HLA-A*29:02", "HLA-A*30:01", "HLA-A*30:02",
"HLA-A*30:03", "HLA-A*30:04", "HLA-A*31:01", "HLA-A*32:01", "HLA-A*32:07", "HLA-A*32:15", "HLA-A*33:01",
"HLA-A*33:03", "HLA-A*66:01", "HLA-A*68:01", "HLA-A*68:02", "HLA-A*68:23", "HLA-A*69:01", "HLA-A*74:01",
"HLA-A*80:01", "HLA-B*07:01", "HLA-B*07:02", "HLA-B*08:01", "HLA-B*08:02", "HLA-B*08:03", "HLA-B*12:01",
"HLA-B*13:02", "HLA-B*14:01", "HLA-B*14:02", "HLA-B*15:01", "HLA-B*15:02", "HLA-B*15:03", "HLA-B*15:08",
"HLA-B*15:09", "HLA-B*15:10", "HLA-B*15:13", "HLA-B*15:16", "HLA-B*15:17", "HLA-B*15:42", "HLA-B*18:01",
"HLA-B*27:01", "HLA-B*27:02", "HLA-B*27:03", "HLA-B*27:04", "HLA-B*27:05", "HLA-B*27:06", "HLA-B*27:09",
"HLA-B*27:10", "HLA-B*27:20", "HLA-B*35:01", "HLA-B*35:02", "HLA-B*35:03", "HLA-B*35:08", "HLA-B*37:01",
"HLA-B*38:01", "HLA-B*39:01", "HLA-B*39:06", "HLA-B*39:09", "HLA-B*39:10", "HLA-B*40:01", "HLA-B*40:02",
"HLA-B*40:13", "HLA-B*41:03", "HLA-B*41:04", "HLA-B*42:01", "HLA-B*42:02", "HLA-B*44:01", "HLA-B*44:02",
"HLA-B*44:03", "HLA-B*44:05", "HLA-B*45:01", "HLA-B*45:06", "HLA-B*46:01", "HLA-B*48:01", "HLA-B*51:01",
"HLA-B*51:02", "HLA-B*52:01", "HLA-B*53:01", "HLA-B*54:01", "HLA-B*55:01", "HLA-B*55:02", "HLA-B*56:01",
"HLA-B*57:01", "HLA-B*57:02", "HLA-B*57:03", "HLA-B*58:01", "HLA-B*58:02", "HLA-B*60:01", "HLA-B*61:01",
"HLA-B*62:01", "HLA-B*73:01", "HLA-B*81:01", "HLA-B*83:01"])
__supported_length = frozenset([5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22])
__name = "mhcnuggets-class-1"
__version = "2.0"
Expand All @@ -123,31 +123,29 @@ def version(self):
# returns the version of the predictor
return self.__version

# the interface defines a function converting Fred2's HLA allele presentation
# into an internal presentation used by different methods.
# for this predictor we won't need it but still have to provide it!
# the function consumes a list of alleles and converts them into the internally used presentation
# Converts FRED2s internal allele representation into the format required by mhcnuggets
def convert_alleles(self, alleles):
# we just use the identity function
return alleles
return ['HLA-' + allele.name.replace('*', '') for allele in alleles]

# additionally the interface defines a function `predict`
# that consumes a list of peptides or a single peptide and optionally a list of allele objects
# this method implements the complete prediction routine
# Converts the internal mhcnuggets-class-1 representation back into a FRED2 representation
def revert_allele_repr(self, allele):
allele = allele.replace('HLA-', '')
allele = allele[:1] + '*' + allele[1:]
return allele

# predicts the binding affinity for a set of peptides and alleles
def predict(self, peptides, alleles=None, binary=False, **kwargs):

# test whether one peptide or a list
if isinstance(peptides, basestring):
peptides = list(peptides)
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.name in self.supportedAlleles, alleles)

result = {}
alleles = filter(lambda a: a in self.supportedAlleles, alleles)

# fetch peptides as strings
peptides = [str(peptide) for peptide in peptides]
Expand All @@ -158,9 +156,12 @@ def predict(self, peptides, alleles=None, binary=False, **kwargs):
for peptide in peptides:
file.write(peptide + "\n")

# predict bindings
alleles = self.convert_alleles(alleles)
result = {}
# predict binding affinities
for a in alleles:
result[a] = {}
allele_repr = self.revert_allele_repr(a)
result[allele_repr] = {}
tmp_output_file = tempfile.NamedTemporaryFile().name
mhcnuggets_predict(class_='I',
peptides_path=tmp_input_file,
Expand All @@ -173,17 +174,18 @@ def predict(self, peptides, alleles=None, binary=False, **kwargs):
# skip header
reader.next()

# assign binding affinities
for row in reader:
content = row[0].split(',')
peptide = content[0]
binding_affinity = content[1]
if binary:
if binding_affinity <= 500:
result[a][peptide] = 1.0
result[allele_repr][peptide] = 1.0
else:
result[a][peptide] = 0.0
result[allele_repr][peptide] = 0.0
else:
result[a][peptide] = binding_affinity
result[allele_repr][peptide] = binding_affinity

# create EpitopePredictionResult object. This is a multi-indexed DataFrame
# with Peptide and Method as multi-index and alleles as columns
Expand Down Expand Up @@ -264,45 +266,50 @@ def version(self):
# returns the version of the predictor
return self.__version

# the interface defines a function converting Fred2's HLA allele presentation
# into an internal presentation used by different methods.
# for this predictor we won't need it but still have to provide it!
# the function consumes a list of alleles and converts them into the internally used presentation
# Converts FRED2s internal allele representation into the format required by mhcnuggets-class-2
def convert_alleles(self, alleles):
return [allele.replace(':', '').replace('*', '') for allele in alleles]
return ['HLA-' + allele.name.replace('*', '') for allele in alleles]

# Converts the internal mhcnuggets-class-2 representation back into a FRED2 representation
def revert_allele_repr(self, allele):
allele = allele.replace('HLA-', '')
# since we need to support single and double mhc2 alleles
allele_split = allele.split('-')
if len(allele_split) > 1:
return allele_split[0][:4] + '*' + allele_split[0][4:] + '-' + allele_split[1][:4] + '*' + allele_split[1][4:]
else:
return allele_split[0][:4] + '*' + allele_split[0][4:]

# additionally the interface defines a function `predict`
# that consumes a list of peptides or a single peptide and optionally a list
# of allele objects
#
# this method implements the complete prediction routine
# predicts the binding affinity for a set of peptides and alleles
def predict(self, peptides, alleles=None, binary=False, **kwargs):

# test whether one peptide or a list
if isinstance(peptides, basestring):
peptides = list(peptides)
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.name in self.supportedAlleles, alleles)

result = {}
alleles = filter(lambda a: a in self.supportedAlleles, alleles)

# fetch peptides as strings
peptides = [str(peptide) for peptide in peptides]

alleles = self.convert_alleles(alleles)

# write peptides temporarily, new line separated
tmp_input_file = tempfile.NamedTemporaryFile().name
with open(tmp_input_file, 'wb') as file:
for peptide in peptides:
file.write(peptide + "\n")

result = {}
# predict bindings
for a in alleles:
result[a] = {}
allele_repr = self.revert_allele_repr(a)
result[allele_repr] = {}
tmp_output_file = tempfile.NamedTemporaryFile().name

mhcnuggets_predict(class_='II',
Expand All @@ -322,11 +329,11 @@ def predict(self, peptides, alleles=None, binary=False, **kwargs):
binding_affinity = content[1]
if binary:
if binding_affinity <= 500:
result[a][peptide] = 1.0
result[allele_repr][peptide] = 1.0
else:
result[a][peptide] = 0.0
result[allele_repr][peptide] = 0.0
else:
result[a][peptide] = binding_affinity
result[allele_repr][peptide] = binding_affinity

# create EpitopePredictionResult object. This is a multi-indexed DataFrame
# with Peptide and Method as multi-index and alleles as columns
Expand Down Expand Up @@ -386,27 +393,31 @@ def version(self):
# returns the version of the predictor
return self.__version

# the interface defines a function converting Fred2's HLA allele presentation
# into an internal presentation used by different methods.
# for this predictor we won't need it but still have to provide it!
# the function consumes a list of alleles and converts them into the internally used presentation
# converts internal FRED2 HLA representations into an internal representation used by MHCFlurry
def convert_alleles(self, alleles):
# we just use the identity function
return alleles

# additionally the interface defines a function `predict`
return ['HLA-' + allele.name.replace(':', '').replace('*', '') for allele in alleles]

# Converts the internal MHCFlurry representation back into a FRED2 representation
def revert_allele_repr(self, allele):
allele = allele.replace('HLA-', '')
allele = allele[:1] + '*' + allele[1:3] + ':' + allele[3:]
return allele

# predicts the binding affinity for a set of peptides and alleles
def predict(self, peptides, alleles=None, binary=False, **kwargs):

# test whether one peptide or a list
if isinstance(peptides, basestring):
peptides = list(peptides)
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.name in self.supportedAlleles, 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'],
Expand All @@ -417,20 +428,21 @@ def predict(self, peptides, alleles=None, binary=False, **kwargs):
# load model
predictor = Class1AffinityPredictor.load()

# predict and assign binding affinities
result = {}

for a in alleles:
result[a] = {}
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])
binding_affinity = predictor.predict(allele=a, peptides=[seq])[0]
if binary:
if binding_affinity <= 500:
result[a][p] = 1.0
result[allele_repr][p] = 1.0
else:
result[a][p] = 0.0
result[allele_repr][p] = 0.0
else:
result[a][p] = binding_affinity
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
Expand Down

0 comments on commit 12fcaf8

Please sign in to comment.