Source code for phages2050.features.extractors.proteins
from typing import Mapping, Union, List, Iterator
from Bio.SeqRecord import SeqRecord
from Bio.SeqUtils.ProtParam import ProteinAnalysis
from Bio.SeqIO.FastaIO import FastaIterator
import pandas as pd
[docs]class ProteinFeatureExtractor:
"""
Feature extraction from protein sequence for
Machine Learning classification or deeper analysis
Example usage:
from features.extractors.proteins import ProteinFeatureExtractor
pfe = ProteinFeatureExtractor(protein_sequence='MAKINELLRESTTTNSNSIGRPNLVALTRATTKLIYSDIVATQRTNQPVAA')
pfe.get_features()
"""
FEATURE_NAMES = [
"protein_length",
"gravy",
"molecular_weight",
"aromaticity",
"instability_index",
"isoelectric_point",
"flexibility",
"mec_cysteines",
"mec_cystines",
"ssf_helix",
"ssf_turn",
"ssf_sheet",
]
def __init__(self, protein_sequence: str):
self.protein_sequence = self._normalize(protein_sequence)
self.protein_analysis = ProteinAnalysis(self.protein_sequence)
@staticmethod
def _normalize(source: Union[str, SeqRecord]) -> str:
"""
Normalize each protein sequence
to uppercase and without blank chars
"""
# If source is a string
if isinstance(source, str):
entry = source
# If source is a BioPython object with seq field
else:
entry = source.seq
return str(entry).upper().strip()
def _get_protein_length(self) -> int:
"""
Protein length
"""
return len(self.protein_sequence)
def _calculate_gravy(self) -> float:
"""
GRAVY (Grand Average of Hydropathy) index score
is calculated by adding the hydropathy value for
each residue and then dividing by the length of
the protein sequence
Negative GRAVY value indicates that the protein
is non-polar and Positive value indicates that
the protein is polar
"""
return self.protein_analysis.gravy()
def _calculate_molecular_weight(self) -> float:
"""
Molecular Weight is calculated as the sum
of atomic masses of all atoms in the molecul
"""
return self.protein_analysis.molecular_weight()
def _calculate_aromaticity(self) -> float:
"""
Aromaticity is used to describe a planar, cyclic
molecule with a ring of resonance bonds which is
more stable when compared to other connective or
geometric arrangements consisting of the same set
of atoms
"""
return self.protein_analysis.aromaticity()
def _calculate_instability_index(self) -> float:
"""
Instability index gives an estimate of the stability
of the protein in a test tube
Any value above 40 means that the protein is unstable
(has a short half life)
"""
return self.protein_analysis.instability_index()
def _calculate_isoelectric_point(self) -> float:
"""
Isoelectric point (pI) is the pH at which net charge of
the protein is zero. Isoelectric point is widely useful
for choosing a buffer system for purification and
crystallisation of a given protein
"""
return self.protein_analysis.isoelectric_point()
def _calculate_flexibility(self) -> float:
"""
Flexibility is of overwhelming importance for protein function,
because of the changes in protein structure during interactions
with binding partners
"""
return sum(self.protein_analysis.flexibility())
def _calculate_molar_extinction_coefficient(self) -> Mapping[str, float]:
"""
Molar extinction coefficient of a protein sequence can be calculated
from the molar extension coefficient of amino acids which are
Cystine, Tyrosine and Tryptophan
"""
cysteines, cystines = self.protein_analysis.molar_extinction_coefficient()
residues = {self.FEATURE_NAMES[7]: cysteines, self.FEATURE_NAMES[8]: cystines}
return residues
def _calculate_secondary_structure_fraction(self) -> Mapping[str, float]:
"""
This function returns a list of the fraction of amino acids which
tend to be in Helix, Turn or Sheet
Amino acids present in Turn are:
Asparagine (N), Proline (P), Glycine (G), Serine (S)
Amino acids present in Sheets are:
Glutamic acid (E), Methionine (M), Alanine (A), Leucine (L)
"""
helix, turn, sheet = self.protein_analysis.secondary_structure_fraction()
fractions = {
self.FEATURE_NAMES[9]: helix,
self.FEATURE_NAMES[10]: turn,
self.FEATURE_NAMES[11]: sheet,
}
return fractions
[docs] def get_features(self) -> Mapping[str, Union[int, float, None]]:
"""
Return full feature space for single protein as Python dict
"""
features = {
self.FEATURE_NAMES[0]: self._get_protein_length(),
self.FEATURE_NAMES[1]: self._calculate_gravy(),
self.FEATURE_NAMES[2]: self._calculate_molecular_weight(),
self.FEATURE_NAMES[3]: self._calculate_aromaticity(),
self.FEATURE_NAMES[4]: self._calculate_instability_index(),
self.FEATURE_NAMES[5]: self._calculate_isoelectric_point(),
self.FEATURE_NAMES[6]: self._calculate_flexibility(),
}
features.update(self._calculate_molar_extinction_coefficient())
features.update(self._calculate_secondary_structure_fraction())
return features
[docs]class MultifastaProteinFeatureExtractor:
"""
Feature extraction from proteins sequences from multifasta file
This class allows you to create DataFrame or save it as CSV
Example usage:
from features.extractors.proteins import MultifastaProteinFeatureExtractor
mpfe = MultifastaProteinFeatureExtractor(protein_sequence='multifasta-example.fasta')
mpfe.to_df()
mpfe.to_csv()
"""
def __init__(self, fasta_path: str):
self.fasta_path = fasta_path
self.entries = self._get_entires()
@staticmethod
def _fasta_reader(filename: str) -> Iterator:
"""
Read FASTA file content including multifasta format
"""
with open(filename) as handle:
for record in FastaIterator(handle):
yield record
def _get_entires(self) -> List:
"""
Extract each entry (protein) from the multifasta
"""
entries = list(self._fasta_reader(self.fasta_path))
return entries
[docs] def to_df(self) -> pd.DataFrame:
"""
Return extracted features from each proteins as DataFrame
"""
data = []
for protein_sequence in self.entries:
protein_features = ProteinFeatureExtractor(
protein_sequence=protein_sequence
).get_features()
data.append(protein_features)
df = pd.DataFrame(data=data, columns=ProteinFeatureExtractor.FEATURE_NAMES)
return df
[docs] def to_csv(self, csv_fname: str) -> None:
"""
Return DataFrame as CSV file
(filename format: <csv_fname>.csv)
"""
df = self.to_df()
df.to_csv(csv_fname, index=False)