def test_get_sprot_raw(self):
"""Bio.ExPASy.get_sprot_raw("O23729")"""
identifier = "O23729"
# This is to catch an error page from our proxy:
handle = UndoHandle(ExPASy.get_sprot_raw(identifier))
if _as_string(handle.peekline()).startswith("<!DOCTYPE HTML"):
raise IOError
record = SeqIO.read(handle, "swiss")
handle.close()
self.assertEqual(record.id, identifier)
self.assertEqual(len(record), 394)
self.assertEqual(seguid(record.seq), "5Y08l+HJRDIlhLKzFEfkcKd1dkM")
def extract_organisms(file, num_records):
scanner = Fasta._Scanner()
consumer = SpeciesExtractor()
file_to_parse = UndoHandle(open(file, "r"))
for fasta_record in range(num_records):
scanner.feed(file_to_parse, consumer)
file_to_parse.close()
return consumer.species_list
def test_get_sprot_raw(self):
"""Bio.ExPASy.get_sprot_raw("O23729")"""
identifier = "O23729"
try:
#This is to catch an error page from our proxy:
handle = UndoHandle(ExPASy.get_sprot_raw(identifier))
if _as_string(handle.peekline()).startswith("<!DOCTYPE HTML"):
raise IOError
record = SeqIO.read(handle, "swiss")
handle.close()
except IOError:
raise MissingExternalDependencyError(
"internet (or maybe just ExPASy) not available")
self.assertEqual(record.id, identifier)
self.assertEqual(len(record), 394)
self.assertEqual(seguid(record.seq), "5Y08l+HJRDIlhLKzFEfkcKd1dkM")
def PdbAtomIterator(handle):
"""Returns SeqRecord objects for each chain in a PDB file
The sequences are derived from the 3D structure (ATOM records), not the
SEQRES lines in the PDB file header.
Unrecognised three letter amino acid codes (e.g. "CSD") from HETATM entries
are converted to "X" in the sequence.
In addition to information from the PDB header (which is the same for all
records), the following chain specific information is placed in the
annotation:
record.annotations["residues"] = List of residue ID strings
record.annotations["chain"] = Chain ID (typically A, B ,...)
record.annotations["model"] = Model ID (typically zero)
Where amino acids are missing from the structure, as indicated by residue
numbering, the sequence is filled in with 'X' characters to match the size
of the missing region, and None is included as the corresponding entry in
the list record.annotations["residues"].
This function uses the Bio.PDB module to do most of the hard work. The
annotation information could be improved but this extra parsing should be
done in parse_pdb_header, not this module.
"""
# Only import PDB when needed, to avoid/delay NumPy dependency in SeqIO
from Bio.PDB import PDBParser
from Bio.SeqUtils import seq1
from Bio.SCOP.three_to_one_dict import to_one_letter_code
def restype(residue):
"""Return a residue's type as a one-letter code.
Non-standard residues (e.g. CSD, ANP) are returned as 'X'.
"""
return seq1(residue.resname, custom_map=to_one_letter_code)
# Deduce the PDB ID from the PDB header
# ENH: or filename?
from Bio.File import UndoHandle
undo_handle = UndoHandle(handle)
firstline = undo_handle.peekline()
if firstline.startswith("HEADER"):
pdb_id = firstline[62:66]
else:
warnings.warn("First line is not a 'HEADER'; can't determine PDB ID")
pdb_id = '????'
struct = PDBParser().get_structure(pdb_id, undo_handle)
model = struct[0]
for chn_id, chain in sorted(model.child_dict.iteritems()):
# HETATM mod. res. policy: remove mod if in sequence, else discard
residues = [res for res in chain.get_unpacked_list()
if seq1(res.get_resname().upper(),
custom_map=to_one_letter_code) != "X"]
if not residues:
continue
# Identify missing residues in the structure
# (fill the sequence with 'X' residues in these regions)
gaps = []
rnumbers = [r.id[1] for r in residues]
for i, rnum in enumerate(rnumbers[:-1]):
if rnumbers[i+1] != rnum + 1:
# It's a gap!
gaps.append((i+1, rnum, rnumbers[i+1]))
if gaps:
res_out = []
prev_idx = 0
for i, pregap, postgap in gaps:
if postgap > pregap:
gapsize = postgap - pregap - 1
res_out.extend(map(restype, residues[prev_idx:i]))
prev_idx = i
res_out.append('X'*gapsize)
# Last segment
res_out.extend(map(restype, residues[prev_idx:]))
else:
warnings.warn("Ignoring out-of-order residues after a gap",
UserWarning)
# Keep the normal part, drop the out-of-order segment
# (presumably modified or hetatm residues, e.g. 3BEG)
res_out.extend(map(restype, residues[prev_idx:i]))
else:
# No gaps
res_out = map(restype, residues)
record_id = "%s:%s" % (pdb_id, chn_id)
# ENH - model number in SeqRecord id if multiple models?
# id = "Chain%s" % str(chain.id)
# if len(structure) > 1 :
# id = ("Model%s|" % str(model.id)) + id
record = SeqRecord(Seq(''.join(res_out), generic_protein),
id=record_id,
description=record_id,
)
# The PDB header was loaded as a dictionary, so let's reuse it all
record.annotations = struct.header.copy()
# Plus some chain specifics:
record.annotations["model"] = model.id
record.annotations["chain"] = chain.id
# Start & end
record.annotations["start"] = int(rnumbers[0])
record.annotations["end"] = int(rnumbers[-1])
# ENH - add letter annotations -- per-residue info, e.g. numbers
yield record
def PdbAtomIterator(handle):
"""Return SeqRecord objects for each chain in a PDB file.
The sequences are derived from the 3D structure (ATOM records), not the
SEQRES lines in the PDB file header.
Unrecognised three letter amino acid codes (e.g. "CSD") from HETATM entries
are converted to "X" in the sequence.
In addition to information from the PDB header (which is the same for all
records), the following chain specific information is placed in the
annotation:
record.annotations["residues"] = List of residue ID strings
record.annotations["chain"] = Chain ID (typically A, B ,...)
record.annotations["model"] = Model ID (typically zero)
Where amino acids are missing from the structure, as indicated by residue
numbering, the sequence is filled in with 'X' characters to match the size
of the missing region, and None is included as the corresponding entry in
the list record.annotations["residues"].
This function uses the Bio.PDB module to do most of the hard work. The
annotation information could be improved but this extra parsing should be
done in parse_pdb_header, not this module.
This gets called internally via Bio.SeqIO for the atom based interpretation
of the PDB file format:
>>> from Bio import SeqIO
>>> for record in SeqIO.parse("PDB/1A8O.pdb", "pdb-atom"):
... print("Record id %s, chain %s" % (record.id, record.annotations["chain"]))
...
Record id 1A8O:A, chain A
Equivalently,
>>> with open("PDB/1A8O.pdb") as handle:
... for record in PdbAtomIterator(handle):
... print("Record id %s, chain %s" % (record.id, record.annotations["chain"]))
...
Record id 1A8O:A, chain A
"""
# TODO - Add record.annotations to the doctest, esp the residues (not working?)
# Only import PDB when needed, to avoid/delay NumPy dependency in SeqIO
from Bio.PDB import PDBParser
# Deduce the PDB ID from the PDB header
# ENH: or filename?
from Bio.File import UndoHandle
undo_handle = UndoHandle(handle)
firstline = undo_handle.peekline()
# check if file is empty
if firstline == '':
raise ValueError("Empty file.")
if firstline.startswith("HEADER"):
pdb_id = firstline[62:66]
else:
warnings.warn("First line is not a 'HEADER'; can't determine PDB ID. "
"Line: %r" % firstline, BiopythonParserWarning)
pdb_id = '????'
struct = PDBParser().get_structure(pdb_id, undo_handle)
for record in AtomIterator(pdb_id, struct):
# The PDB header was loaded as a dictionary, so let's reuse it all
record.annotations.update(struct.header)
# ENH - add letter annotations -- per-residue info, e.g. numbers
yield record
def __init__(self, handle, __parse_hit_table=False):
"""Initialize the class."""
self.handle = UndoHandle(handle)
self._preamble = self._parse_preamble()
class FastaM10Parser(object):
"""Parser for Bill Pearson's FASTA suite's -m 10 output."""
def __init__(self, handle, __parse_hit_table=False):
self.handle = UndoHandle(handle)
self._preamble = self._parse_preamble()
def __iter__(self):
for qresult in self._parse_qresult():
# re-set desc, for hsp query description
qresult.description = qresult.description
yield qresult
def _parse_preamble(self):
"""Parses the Fasta preamble for Fasta flavor and version."""
preamble = {}
while True:
self.line = self.handle.readline()
# this should be the line just before the first qresult
if self.line.startswith('Query'):
break
# try to match for version line
elif self.line.startswith(' version'):
preamble['version'] = self.line.split(' ')[2]
else:
# try to match for flavor line
flav_match = re.match(_RE_FLAVS, self.line.lower())
if flav_match:
preamble['program'] = flav_match.group(0)
return preamble
def __parse_hit_table(self):
"""Parses hit table rows."""
# move to the first row
self.line = self.handle.readline()
# parse hit table until we see an empty line
hit_rows = []
while self.line and not self.line.strip():
hit_rows.append(self.line.strip())
self.line = self.handle.readline()
return hit_rows
def _parse_qresult(self):
# initial qresult value
qresult = None
hit_rows = []
# state values
state_QRES_NEW = 1
state_QRES_HITTAB = 3
state_QRES_CONTENT = 5
state_QRES_END = 7
while True:
# one line before the hit table
if self.line.startswith('The best scores are:'):
qres_state = state_QRES_HITTAB
# the end of a query or the file altogether
elif self.line.strip() == '>>>///' or not self.line:
qres_state = state_QRES_END
# the beginning of a new query
elif not self.line.startswith('>>>') and '>>>' in self.line:
qres_state = state_QRES_NEW
# the beginning of the query info and its hits + hsps
elif self.line.startswith('>>>') and not \
self.line.strip() == '>>><<<':
qres_state = state_QRES_CONTENT
# default qres mark
else:
qres_state = None
if qres_state is not None:
if qres_state == state_QRES_HITTAB:
# parse hit table if flag is set
hit_rows = self.__parse_hit_table()
elif qres_state == state_QRES_END:
yield _set_qresult_hits(qresult, hit_rows)
break
elif qres_state == state_QRES_NEW:
# if qresult is filled, yield it first
if qresult is not None:
yield _set_qresult_hits(qresult, hit_rows)
regx = re.search(_RE_ID_DESC_SEQLEN, self.line)
query_id = regx.group(1)
seq_len = regx.group(3)
desc = regx.group(2)
qresult = QueryResult(id=query_id)
qresult.seq_len = int(seq_len)
# get target from the next line
self.line = self.handle.readline()
qresult.target = [x for x in self.line.split(' ')
if x][1].strip()
if desc is not None:
qresult.description = desc
# set values from preamble
for key, value in self._preamble.items():
setattr(qresult, key, value)
elif qres_state == state_QRES_CONTENT:
assert self.line[3:].startswith(qresult.id), self.line
for hit, strand in self._parse_hit(query_id):
# HACK: re-set desc, for hsp hit and query description
hit.description = hit.description
hit.query_description = qresult.description
# if hit is not in qresult, append it
if hit.id not in qresult:
qresult.append(hit)
# otherwise, it might be the same hit with a different strand
else:
# make sure strand is different and then append hsp to
# existing hit
for hsp in hit.hsps:
assert strand != hsp.query_strand
qresult[hit.id].append(hsp)
self.line = self.handle.readline()
def _parse_hit(self, query_id):
while True:
self.line = self.handle.readline()
if self.line.startswith('>>'):
break
state = _STATE_NONE
strand = None
hsp_list = []
while True:
peekline = self.handle.peekline()
# yield hit if we've reached the start of a new query or
# the end of the search
if peekline.strip() in [">>><<<", ">>>///"] or \
(not peekline.startswith('>>>') and '>>>' in peekline):
# append last parsed_hsp['hit']['seq'] line
if state == _STATE_HIT_BLOCK:
parsed_hsp['hit']['seq'] += self.line.strip()
elif state == _STATE_CONS_BLOCK:
hsp.aln_annotation['similarity'] += \
self.line.strip('\r\n')
# process HSP alignment and coordinates
_set_hsp_seqs(hsp, parsed_hsp, self._preamble['program'])
hit = Hit(hsp_list)
hit.description = hit_desc
hit.seq_len = seq_len
yield hit, strand
hsp_list = []
break
# yield hit and create a new one if we're still in the same query
elif self.line.startswith('>>'):
# try yielding, if we have hsps
if hsp_list:
_set_hsp_seqs(hsp, parsed_hsp, self._preamble['program'])
hit = Hit(hsp_list)
hit.description = hit_desc
hit.seq_len = seq_len
yield hit, strand
hsp_list = []
# try to get the hit id and desc, and handle cases without descs
try:
hit_id, hit_desc = self.line[2:].strip().split(' ', 1)
except ValueError:
hit_id = self.line[2:].strip().split(' ', 1)[0]
hit_desc = ''
# create the HSP object for Hit
frag = HSPFragment(hit_id, query_id)
hsp = HSP([frag])
hsp_list.append(hsp)
# set or reset the state to none
state = _STATE_NONE
parsed_hsp = {'query': {}, 'hit': {}}
# create and append a new HSP if line starts with '>--'
elif self.line.startswith('>--'):
# set seq attributes of previous hsp
_set_hsp_seqs(hsp, parsed_hsp, self._preamble['program'])
# and create a new one
frag = HSPFragment(hit_id, query_id)
hsp = HSP([frag])
hsp_list.append(hsp)
# set the state ~ none yet
state = _STATE_NONE
parsed_hsp = {'query': {}, 'hit': {}}
# this is either query or hit data in the HSP, depending on the state
elif self.line.startswith('>'):
if state == _STATE_NONE:
# make sure it's the correct query
assert query_id.startswith(self.line[1:].split(' ')[0]), \
"%r vs %r" % (query_id, self.line)
state = _STATE_QUERY_BLOCK
parsed_hsp['query']['seq'] = ''
elif state == _STATE_QUERY_BLOCK:
# make sure it's the correct hit
assert hit_id.startswith(self.line[1:].split(' ')[0])
state = _STATE_HIT_BLOCK
parsed_hsp['hit']['seq'] = ''
# check for conservation block
elif self.line.startswith('; al_cons'):
state = _STATE_CONS_BLOCK
hsp.fragment.aln_annotation['similarity'] = ''
elif self.line.startswith(';'):
# Fasta outputs do not make a clear distinction between Hit
# and HSPs, so we check the attribute names to determine
# whether it belongs to a Hit or HSP
regx = re.search(_RE_ATTR, self.line.strip())
name = regx.group(1)
value = regx.group(2)
# for values before the '>...' query block
if state == _STATE_NONE:
if name in _HSP_ATTR_MAP:
attr_name, caster = _HSP_ATTR_MAP[name]
if caster is not str:
value = caster(value)
if name in ['_ident', '_sim']:
value *= 100
setattr(hsp, attr_name, value)
# otherwise, pool the values for processing later
elif state == _STATE_QUERY_BLOCK:
parsed_hsp['query'][name] = value
elif state == _STATE_HIT_BLOCK:
if name == '_len':
seq_len = int(value)
else:
parsed_hsp['hit'][name] = value
# for values in the hit block
else:
raise ValueError("Unexpected line: %r" % self.line)
# otherwise, it must be lines containing the sequences
else:
assert '>' not in self.line
# if we're in hit, parse into hsp.hit
if state == _STATE_HIT_BLOCK:
parsed_hsp['hit']['seq'] += self.line.strip()
elif state == _STATE_QUERY_BLOCK:
parsed_hsp['query']['seq'] += self.line.strip()
elif state == _STATE_CONS_BLOCK:
hsp.fragment.aln_annotation['similarity'] += \
self.line.strip('\r\n')
# we should not get here!
else:
raise ValueError("Unexpected line: %r" % self.line)
self.line = self.handle.readline()
from Bio.MEME import Parser
from Bio import ParserSupport
from Bio.File import UndoHandle
meme_tests = [
"meme.dna.oops.txt", "meme.protein.oops.txt", "meme.protein.tcm.txt"
]
mast_tests = [
"mast.dna.oops.txt", "mast.protein.oops.txt", "mast.protein.tcm.txt"
]
print "Testing MEME Scanner"
datafile = os.path.join("MEME", meme_tests[0])
uhandle = UndoHandle(open(datafile))
scanner = Parser._MEMEScanner()
consumer = ParserSupport.TaggingConsumer()
scanner.feed(uhandle, consumer)
print "Running tests on MEME parser"
meme_parser = Parser.MEMEParser()
for test in meme_tests:
print "*" * 50, "TESTING %s" % test
datafile = os.path.join("MEME", test)
rec = meme_parser.parse(open(datafile))
print "Testing MEME Scanner"
def __init__(self, filename):
SearchIndexer.__init__(self, filename)
self._handle = UndoHandle(self._handle)
def __init__(self, handle, __parse_hit_table=False):
self.handle = UndoHandle(handle)
self._preamble = self._parse_preamble()
def PdbAtomIterator(handle):
"""Return SeqRecord objects for each chain in a PDB file.
The sequences are derived from the 3D structure (ATOM records), not the
SEQRES lines in the PDB file header.
Unrecognised three letter amino acid codes (e.g. "CSD") from HETATM entries
are converted to "X" in the sequence.
In addition to information from the PDB header (which is the same for all
records), the following chain specific information is placed in the
annotation:
record.annotations["residues"] = List of residue ID strings
record.annotations["chain"] = Chain ID (typically A, B ,...)
record.annotations["model"] = Model ID (typically zero)
Where amino acids are missing from the structure, as indicated by residue
numbering, the sequence is filled in with 'X' characters to match the size
of the missing region, and None is included as the corresponding entry in
the list record.annotations["residues"].
This function uses the Bio.PDB module to do most of the hard work. The
annotation information could be improved but this extra parsing should be
done in parse_pdb_header, not this module.
This gets called internally via Bio.SeqIO for the atom based interpretation
of the PDB file format:
>>> from Bio import SeqIO
>>> for record in SeqIO.parse("PDB/1A8O.pdb", "pdb-atom"):
... print("Record id %s, chain %s" % (record.id, record.annotations["chain"]))
...
Record id 1A8O:A, chain A
Equivalently,
>>> with open("PDB/1A8O.pdb") as handle:
... for record in PdbAtomIterator(handle):
... print("Record id %s, chain %s" % (record.id, record.annotations["chain"]))
...
Record id 1A8O:A, chain A
"""
# TODO - Add record.annotations to the doctest, esp the residues (not working?)
# Only import PDB when needed, to avoid/delay NumPy dependency in SeqIO
from Bio.PDB import PDBParser
# Deduce the PDB ID from the PDB header
# ENH: or filename?
from Bio.File import UndoHandle
undo_handle = UndoHandle(handle)
firstline = undo_handle.peekline()
if firstline.startswith("HEADER"):
pdb_id = firstline[62:66]
else:
warnings.warn(
"First line is not a 'HEADER'; can't determine PDB ID. "
"Line: %r" % firstline, BiopythonParserWarning)
pdb_id = '????'
struct = PDBParser().get_structure(pdb_id, undo_handle)
for record in AtomIterator(pdb_id, struct):
# The PDB header was loaded as a dictionary, so let's reuse it all
record.annotations.update(struct.header)
# ENH - add letter annotations -- per-residue info, e.g. numbers
yield record
def __init__(self, filename):
"""Initialize the class."""
SearchIndexer.__init__(self, filename)
self._handle = UndoHandle(self._handle)
def __init__(self, handle, __parse_hit_table=False):
"""Initialize the class."""
self.handle = UndoHandle(handle)
self._preamble = self._parse_preamble()
def CifSeqresIterator(handle):
"""Return SeqRecord objects for each chain in an mmCIF file.
The sequences are derived from the _entity_poly_seq entries in the mmCIF
file, not the atoms of the 3D structure.
Specifically, these mmCIF records are handled: _pdbx_poly_seq_scheme and
_struct_ref_seq. The _pdbx_poly_seq records contain sequence information,
and the _struct_ref_seq records contain database cross-references.
See:
http://mmcif.wwpdb.org/dictionaries/mmcif_pdbx_v40.dic/Categories/pdbx_poly_seq_scheme.html
and
http://mmcif.wwpdb.org/dictionaries/mmcif_pdbx_v50.dic/Categories/struct_ref_seq.html
This gets called internally via Bio.SeqIO for the sequence-based
interpretation of the mmCIF file format:
>>> from Bio import SeqIO
>>> for record in SeqIO.parse("PDB/1A8O.cif", "cif-seqres"):
... print("Record id %s, chain %s" % (record.id, record.annotations["chain"]))
... print(record.dbxrefs)
...
Record id 1A8O:A, chain A
['UNP:P12497', 'UNP:POL_HV1N5']
Equivalently,
>>> with open("PDB/1A8O.cif") as handle:
... for record in CifSeqresIterator(handle):
... print("Record id %s, chain %s" % (record.id, record.annotations["chain"]))
... print(record.dbxrefs)
...
Record id 1A8O:A, chain A
['UNP:P12497', 'UNP:POL_HV1N5']
Note the chain is recorded in the annotations dictionary, and any mmCIF
_struct_ref_seq entries are recorded in the database cross-references list.
"""
# Late-binding import to avoid circular dependency on SeqIO in Bio.SeqUtils
from Bio.SeqUtils import seq1
# Only import PDB when needed, to avoid/delay NumPy dependency in SeqIO
from Bio.PDB.MMCIF2Dict import MMCIF2Dict
from Bio.File import UndoHandle
undo_handle = UndoHandle(handle)
# check if file is empty
if not undo_handle.peekline():
raise ValueError("Empty file.")
chains = collections.defaultdict(list)
metadata = collections.defaultdict(list)
records = MMCIF2Dict(undo_handle)
# Explicitly convert records to list (See #1533).
# If an item is not present, use an empty list
for field in (
PDBX_POLY_SEQ_SCHEME_FIELDS + STRUCT_REF_SEQ_FIELDS + STRUCT_REF_FIELDS
):
if field not in records:
records[field] = []
elif not isinstance(records[field], list):
records[field] = [records[field]]
for asym_id, mon_id in zip(
records["_pdbx_poly_seq_scheme.asym_id"],
records["_pdbx_poly_seq_scheme.mon_id"],
):
mon_id_1l = seq1(mon_id, custom_map=protein_letters_3to1)
chains[asym_id].append(mon_id_1l)
# Build a dict of _struct_ref records, indexed by the id field:
struct_refs = {}
for fields in zip(
records["_struct_ref.id"],
records["_struct_ref.db_name"],
records["_struct_ref.db_code"],
records["_struct_ref.pdbx_db_accession"],
):
ref_id, db_name, db_code, db_acc = fields
struct_refs[ref_id] = {
"database": db_name,
"db_id_code": db_code,
"db_acc": db_acc,
}
# Look through _struct_ref_seq records, look up the corresponding
# _struct_ref and add an entry to the metadata list for this chain.
for fields in zip(
records["_struct_ref_seq.ref_id"],
records["_struct_ref_seq.pdbx_PDB_id_code"],
records["_struct_ref_seq.pdbx_strand_id"],
):
ref_id, pdb_id, chain_id = fields
struct_ref = struct_refs[ref_id]
# The names here mirror those in PdbIO
metadata[chain_id].append({"pdb_id": pdb_id})
metadata[chain_id][-1].update(struct_ref)
for chn_id, residues in sorted(chains.items()):
record = SeqRecord(Seq("".join(residues), generic_protein))
record.annotations = {"chain": chn_id}
if chn_id in metadata:
m = metadata[chn_id][0]
record.id = record.name = "%s:%s" % (m["pdb_id"], chn_id)
record.description = "%s:%s %s" % (
m["database"],
m["db_acc"],
m["db_id_code"],
)
for melem in metadata[chn_id]:
record.dbxrefs.extend(
[
"%s:%s" % (melem["database"], melem["db_acc"]),
"%s:%s" % (melem["database"], melem["db_id_code"]),
]
)
else:
record.id = chn_id
yield record
class FastaM10Parser(object):
"""Parser for Bill Pearson's FASTA suite's -m 10 output."""
def __init__(self, handle, __parse_hit_table=False):
self.handle = UndoHandle(handle)
self._preamble = self._parse_preamble()
def __iter__(self):
for qresult in self._parse_qresult():
# re-set desc, for hsp query description
qresult.description = qresult.description
yield qresult
def _parse_preamble(self):
"""Parses the Fasta preamble for Fasta flavor and version."""
preamble = {}
while True:
self.line = self.handle.readline()
# this should be the line just before the first qresult
if self.line.startswith('Query'):
break
# try to match for version line
elif self.line.startswith(' version'):
preamble['version'] = self.line.split(' ')[2]
else:
# try to match for flavor line
flav_match = re.match(_RE_FLAVS, self.line.lower())
if flav_match:
preamble['program'] = flav_match.group(0)
return preamble
def __parse_hit_table(self):
"""Parses hit table rows."""
# move to the first row
self.line = self.handle.readline()
# parse hit table until we see an empty line
hit_rows = []
while self.line and not self.line.strip():
hit_rows.append(self.line.strip())
self.line = self.handle.readline()
return hit_rows
def _parse_qresult(self):
# initial qresult value
qresult = None
hit_rows = []
# state values
state_QRES_NEW = 1
state_QRES_HITTAB = 3
state_QRES_CONTENT = 5
state_QRES_END = 7
while True:
# one line before the hit table
if self.line.startswith('The best scores are:'):
qres_state = state_QRES_HITTAB
# the end of a query or the file altogether
elif self.line.strip() == '>>>///' or not self.line:
qres_state = state_QRES_END
# the beginning of a new query
elif not self.line.startswith('>>>') and '>>>' in self.line:
qres_state = state_QRES_NEW
# the beginning of the query info and its hits + hsps
elif self.line.startswith('>>>') and not \
self.line.strip() == '>>><<<':
qres_state = state_QRES_CONTENT
# default qres mark
else:
qres_state = None
if qres_state is not None:
if qres_state == state_QRES_HITTAB:
# parse hit table if flag is set
hit_rows = self.__parse_hit_table()
elif qres_state == state_QRES_END:
yield _set_qresult_hits(qresult, hit_rows)
break
elif qres_state == state_QRES_NEW:
# if qresult is filled, yield it first
if qresult is not None:
yield _set_qresult_hits(qresult, hit_rows)
regx = re.search(_RE_ID_DESC_SEQLEN, self.line)
query_id = regx.group(1)
seq_len = regx.group(3)
desc = regx.group(2)
qresult = QueryResult(id=query_id)
qresult.seq_len = int(seq_len)
# get target from the next line
self.line = self.handle.readline()
qresult.target = [x for x in self.line.split(' ') if x][1].strip()
if desc is not None:
qresult.description = desc
# set values from preamble
for key, value in self._preamble.items():
setattr(qresult, key, value)
elif qres_state == state_QRES_CONTENT:
assert self.line[3:].startswith(qresult.id), self.line
for hit, strand in self._parse_hit(query_id):
# HACK: re-set desc, for hsp hit and query description
hit.description = hit.description
hit.query_description = qresult.description
# if hit is not in qresult, append it
if hit.id not in qresult:
qresult.append(hit)
# otherwise, it might be the same hit with a different strand
else:
# make sure strand is different and then append hsp to
# existing hit
for hsp in hit.hsps:
assert strand != hsp.query_strand
qresult[hit.id].append(hsp)
self.line = self.handle.readline()
def _parse_hit(self, query_id):
while True:
self.line = self.handle.readline()
if self.line.startswith('>>'):
break
state = _STATE_NONE
strand = None
hsp_list = []
while True:
peekline = self.handle.peekline()
# yield hit if we've reached the start of a new query or
# the end of the search
if peekline.strip() in [">>><<<", ">>>///"] or \
(not peekline.startswith('>>>') and '>>>' in peekline):
# append last parsed_hsp['hit']['seq'] line
if state == _STATE_HIT_BLOCK:
parsed_hsp['hit']['seq'] += self.line.strip()
elif state == _STATE_CONS_BLOCK:
hsp.aln_annotation['similarity'] += \
self.line.strip('\r\n')
# process HSP alignment and coordinates
_set_hsp_seqs(hsp, parsed_hsp, self._preamble['program'])
hit = Hit(hsp_list)
hit.description = hit_desc
hit.seq_len = seq_len
yield hit, strand
hsp_list = []
break
# yield hit and create a new one if we're still in the same query
elif self.line.startswith('>>'):
# try yielding, if we have hsps
if hsp_list:
_set_hsp_seqs(hsp, parsed_hsp, self._preamble['program'])
hit = Hit(hsp_list)
hit.description = hit_desc
hit.seq_len = seq_len
yield hit, strand
hsp_list = []
# try to get the hit id and desc, and handle cases without descs
try:
hit_id, hit_desc = self.line[2:].strip().split(' ', 1)
except ValueError:
hit_id = self.line[2:].strip().split(' ', 1)[0]
hit_desc = ''
# create the HSP object for Hit
frag = HSPFragment(hit_id, query_id)
hsp = HSP([frag])
hsp_list.append(hsp)
# set or reset the state to none
state = _STATE_NONE
parsed_hsp = {'query': {}, 'hit': {}}
# create and append a new HSP if line starts with '>--'
elif self.line.startswith('>--'):
# set seq attributes of previous hsp
_set_hsp_seqs(hsp, parsed_hsp, self._preamble['program'])
# and create a new one
frag = HSPFragment(hit_id, query_id)
hsp = HSP([frag])
hsp_list.append(hsp)
# set the state ~ none yet
state = _STATE_NONE
parsed_hsp = {'query': {}, 'hit': {}}
# this is either query or hit data in the HSP, depending on the state
elif self.line.startswith('>'):
if state == _STATE_NONE:
# make sure it's the correct query
assert query_id.startswith(self.line[1:].split(' ')[0]), \
"%r vs %r" % (query_id, self.line)
state = _STATE_QUERY_BLOCK
parsed_hsp['query']['seq'] = ''
elif state == _STATE_QUERY_BLOCK:
# make sure it's the correct hit
assert hit_id.startswith(self.line[1:].split(' ')[0])
state = _STATE_HIT_BLOCK
parsed_hsp['hit']['seq'] = ''
# check for conservation block
elif self.line.startswith('; al_cons'):
state = _STATE_CONS_BLOCK
hsp.fragment.aln_annotation['similarity'] = ''
elif self.line.startswith(';'):
# Fasta outputs do not make a clear distinction between Hit
# and HSPs, so we check the attribute names to determine
# whether it belongs to a Hit or HSP
regx = re.search(_RE_ATTR, self.line.strip())
name = regx.group(1)
value = regx.group(2)
# for values before the '>...' query block
if state == _STATE_NONE:
if name in _HSP_ATTR_MAP:
attr_name, caster = _HSP_ATTR_MAP[name]
if caster is not str:
value = caster(value)
if name in ['_ident', '_sim']:
value *= 100
setattr(hsp, attr_name, value)
# otherwise, pool the values for processing later
elif state == _STATE_QUERY_BLOCK:
parsed_hsp['query'][name] = value
elif state == _STATE_HIT_BLOCK:
if name == '_len':
seq_len = int(value)
else:
parsed_hsp['hit'][name] = value
# for values in the hit block
else:
raise ValueError("Unexpected line: %r" % self.line)
# otherwise, it must be lines containing the sequences
else:
assert '>' not in self.line
# if we're in hit, parse into hsp.hit
if state == _STATE_HIT_BLOCK:
parsed_hsp['hit']['seq'] += self.line.strip()
elif state == _STATE_QUERY_BLOCK:
parsed_hsp['query']['seq'] += self.line.strip()
elif state == _STATE_CONS_BLOCK:
hsp.fragment.aln_annotation['similarity'] += \
self.line.strip('\r\n')
# we should not get here!
else:
raise ValueError("Unexpected line: %r" % self.line)
self.line = self.handle.readline()
def __init__(self, handle, __parse_hit_table=False):
self.handle = UndoHandle(handle)
self._preamble = self._parse_preamble()
def PdbAtomIterator(handle):
"""Returns SeqRecord objects for each chain in a PDB file
The sequences are derived from the 3D structure (ATOM records), not the
SEQRES lines in the PDB file header.
Unrecognised three letter amino acid codes (e.g. "CSD") from HETATM entries
are converted to "X" in the sequence.
In addition to information from the PDB header (which is the same for all
records), the following chain specific information is placed in the
annotation:
record.annotations["residues"] = List of residue ID strings
record.annotations["chain"] = Chain ID (typically A, B ,...)
record.annotations["model"] = Model ID (typically zero)
Where amino acids are missing from the structure, as indicated by residue
numbering, the sequence is filled in with 'X' characters to match the size
of the missing region, and None is included as the corresponding entry in
the list record.annotations["residues"].
This function uses the Bio.PDB module to do most of the hard work. The
annotation information could be improved but this extra parsing should be
done in parse_pdb_header, not this module.
This gets called internally via Bio.SeqIO for the atom based interpretation
of the PDB file format:
>>> from Bio import SeqIO
>>> for record in SeqIO.parse("PDB/1A8O.pdb", "pdb-atom"):
... print("Record id %s, chain %s" % (record.id, record.annotations["chain"]))
...
Record id 1A8O:A, chain A
Equivalently,
>>> with open("PDB/1A8O.pdb") as handle:
... for record in PdbAtomIterator(handle):
... print("Record id %s, chain %s" % (record.id, record.annotations["chain"]))
...
Record id 1A8O:A, chain A
"""
# TODO - Add record.annotations to the doctest, esp the residues (not working?)
# Only import PDB when needed, to avoid/delay NumPy dependency in SeqIO
from Bio.PDB import PDBParser
from Bio.SeqUtils import seq1
def restype(residue):
"""Return a residue's type as a one-letter code.
Non-standard residues (e.g. CSD, ANP) are returned as 'X'.
"""
return seq1(residue.resname, custom_map=protein_letters_3to1)
# Deduce the PDB ID from the PDB header
# ENH: or filename?
from Bio.File import UndoHandle
undo_handle = UndoHandle(handle)
firstline = undo_handle.peekline()
if firstline.startswith("HEADER"):
pdb_id = firstline[62:66]
else:
warnings.warn("First line is not a 'HEADER'; can't determine PDB ID")
pdb_id = '????'
struct = PDBParser().get_structure(pdb_id, undo_handle)
model = struct[0]
for chn_id, chain in sorted(model.child_dict.items()):
# HETATM mod. res. policy: remove mod if in sequence, else discard
residues = [
res for res in chain.get_unpacked_list()
if seq1(res.get_resname().upper(), custom_map=protein_letters_3to1)
!= "X"
]
if not residues:
continue
# Identify missing residues in the structure
# (fill the sequence with 'X' residues in these regions)
gaps = []
rnumbers = [r.id[1] for r in residues]
for i, rnum in enumerate(rnumbers[:-1]):
if rnumbers[i + 1] != rnum + 1:
# It's a gap!
gaps.append((i + 1, rnum, rnumbers[i + 1]))
if gaps:
res_out = []
prev_idx = 0
for i, pregap, postgap in gaps:
if postgap > pregap:
gapsize = postgap - pregap - 1
res_out.extend(restype(x) for x in residues[prev_idx:i])
prev_idx = i
res_out.append('X' * gapsize)
else:
warnings.warn("Ignoring out-of-order residues after a gap",
UserWarning)
# Keep the normal part, drop the out-of-order segment
# (presumably modified or hetatm residues, e.g. 3BEG)
res_out.extend(restype(x) for x in residues[prev_idx:i])
break
else:
# Last segment
res_out.extend(restype(x) for x in residues[prev_idx:])
else:
# No gaps
res_out = [restype(x) for x in residues]
record_id = "%s:%s" % (pdb_id, chn_id)
# ENH - model number in SeqRecord id if multiple models?
# id = "Chain%s" % str(chain.id)
# if len(structure) > 1 :
# id = ("Model%s|" % str(model.id)) + id
record = SeqRecord(
Seq(''.join(res_out), generic_protein),
id=record_id,
description=record_id,
)
# The PDB header was loaded as a dictionary, so let's reuse it all
record.annotations = struct.header.copy()
# Plus some chain specifics:
record.annotations["model"] = model.id
record.annotations["chain"] = chain.id
# Start & end
record.annotations["start"] = int(rnumbers[0])
record.annotations["end"] = int(rnumbers[-1])
# ENH - add letter annotations -- per-residue info, e.g. numbers
yield record
