From 30f4442aeded4d9ca376e2d4534de253b7480ee8 Mon Sep 17 00:00:00 2001
From: Tim O'Donnell <timodonnell@gmail.com>
Date: Tue, 9 Apr 2019 22:52:32 -0400
Subject: [PATCH] update encodable_sequences
---
mhcflurry/encodable_sequences.py | 131 +++++++++++++++++--------------
1 file changed, 71 insertions(+), 60 deletions(-)
diff --git a/mhcflurry/encodable_sequences.py b/mhcflurry/encodable_sequences.py
index 47d62f88..2560a8ef 100644
--- a/mhcflurry/encodable_sequences.py
+++ b/mhcflurry/encodable_sequences.py
@@ -132,7 +132,12 @@ class EncodableSequences(object):
@classmethod
def sequences_to_fixed_length_index_encoded_array(
- klass, sequences, left_edge=4, right_edge=4, max_length=15):
+ klass,
+ sequences,
+ alignment_method="pad_middle",
+ left_edge=4,
+ right_edge=4,
+ max_length=15):
"""
Transform a sequence of strings, where each string is of length at least
left_edge + right_edge and at most max_length into strings of length
@@ -161,63 +166,69 @@ class EncodableSequences(object):
-------
numpy array of shape (len(sequences), max_length) and dtype int
"""
-
- # Result array is int32, filled with X (null amino acid) value.
- result = numpy.full(
- fill_value=amino_acid.AMINO_ACID_INDEX['X'],
- shape=(len(sequences), max_length),
- dtype="int32")
-
- df = pandas.DataFrame({"peptide": sequences})
- df["length"] = df.peptide.str.len()
-
- middle_length = max_length - left_edge - right_edge
-
- # For efficiency we handle each supported peptide length using bulk
- # array operations.
- for (length, sub_df) in df.groupby("length"):
- if length < left_edge + right_edge:
- raise ValueError(
- "Sequence '%s' (length %d) unsupported: length must be at "
- "least %d. There are %d total peptides with this length." % (
- sub_df.iloc[0].peptide, length, left_edge + right_edge,
- len(sub_df)))
- if length > max_length:
- raise ValueError(
- "Sequence '%s' (length %d) unsupported: length must be at "
- "most %d. There are %d total peptides with this length." % (
- sub_df.iloc[0].peptide, length, max_length,
- len(sub_df)))
-
- # Array of shape (num peptides, length) giving fixed-length amino
- # acid encoding each peptide of the current length.
- fixed_length_sequences = numpy.stack(
- sub_df.peptide.map(
- lambda s: numpy.array([
- amino_acid.AMINO_ACID_INDEX[char] for char in s
- ])).values)
-
- num_null = max_length - length
- num_null_left = int(math.ceil(num_null / 2))
- num_middle_filled = middle_length - num_null
- middle_start = left_edge + num_null_left
-
- # Set left edge
- result[sub_df.index, :left_edge] = fixed_length_sequences[
- :, :left_edge
- ]
-
- # Set middle.
- result[
- sub_df.index,
- middle_start : middle_start + num_middle_filled
- ] = fixed_length_sequences[
- :, left_edge : left_edge + num_middle_filled
- ]
-
- # Set right edge.
- result[
- sub_df.index,
- -right_edge:
- ] = fixed_length_sequences[:, -right_edge:]
+ result = None
+ if alignment_method == 'pad_middle':
+ # Result array is int32, filled with X (null amino acid) value.
+ result = numpy.full(
+ fill_value=amino_acid.AMINO_ACID_INDEX['X'],
+ shape=(len(sequences), max_length),
+ dtype="int32")
+
+ df = pandas.DataFrame({"peptide": sequences})
+ df["length"] = df.peptide.str.len()
+
+ middle_length = max_length - left_edge - right_edge
+
+ # For efficiency we handle each supported peptide length using bulk
+ # array operations.
+ for (length, sub_df) in df.groupby("length"):
+ if length < left_edge + right_edge:
+ raise ValueError(
+ "Sequence '%s' (length %d) unsupported: length must be at "
+ "least %d. There are %d total peptides with this length." % (
+ sub_df.iloc[0].peptide, length, left_edge + right_edge,
+ len(sub_df)))
+ if length > max_length:
+ raise ValueError(
+ "Sequence '%s' (length %d) unsupported: length must be at "
+ "most %d. There are %d total peptides with this length." % (
+ sub_df.iloc[0].peptide, length, max_length,
+ len(sub_df)))
+
+ # Array of shape (num peptides, length) giving fixed-length amino
+ # acid encoding each peptide of the current length.
+ fixed_length_sequences = numpy.stack(
+ sub_df.peptide.map(
+ lambda s: numpy.array([
+ amino_acid.AMINO_ACID_INDEX[char] for char in s
+ ])).values)
+
+ num_null = max_length - length
+ num_null_left = int(math.ceil(num_null / 2))
+ num_middle_filled = middle_length - num_null
+ middle_start = left_edge + num_null_left
+
+ # Set left edge
+ result[sub_df.index, :left_edge] = fixed_length_sequences[
+ :, :left_edge
+ ]
+
+ # Set middle.
+ result[
+ sub_df.index,
+ middle_start : middle_start + num_middle_filled
+ ] = fixed_length_sequences[
+ :, left_edge : left_edge + num_middle_filled
+ ]
+
+ # Set right edge.
+ result[
+ sub_df.index,
+ -right_edge:
+ ] = fixed_length_sequences[:, -right_edge:]
+ elif alignment_method == "left_pad_right_pad":
+ raise NotImplementedError
+ else:
+ raise NotImplementedError(
+ "Unsupported alignment method: %s" % alignment_method)
return result
--
GitLab