def refineMSA(msa, label=None, rowocc=None, seqid=None, colocc=None, **kwargs):
"""Refine *msa* by removing sequences (rows) and residues (columns) that
contain gaps.
:arg msa: multiple sequence alignment
:type msa: :class:`.MSA`
:arg label: remove columns that are gaps in the sequence matching label,
``msa.getIndex(label)`` must return a sequence index, a PDB identifier
is also acceptable
:type label: str
:arg rowocc: row occupancy, sequences with less occupancy will be
removed after *label* refinement is applied
:type rowocc: float
:arg seqid: keep unique sequences at specified sequence identity level,
unique sequences are identified using :func:`.uniqueSequences`
:type seqid: float
:arg colocc: column occupancy, residue positions with less occupancy
will be removed after other refinements are applied
:type colocc: float
:arg keep: keep columns corresponding to residues not resolved in the PDB
structure, default is **False**, applies when *label* is a PDB
identifier
:arg type: bool
For Pfam MSA data, *label* is UniProt entry name for the protein. You may
also use PDB structure and chain identifiers, e.g. ``'1p38'`` or
``'1p38A'``, for *label* argument and UniProt entry names will be parsed
using :func:`.parsePDBHeader` function (see also :class:`.Polymer` and
:class:`.DBRef`).
The order of refinements are applied in the order of arguments. If *label*
and *unique* is specified is specified, sequence matching *label* will
be kept in the refined :class:`.MSA` although it may be similar to some
other sequence."""
# if msa is a char array, it will be refined but label won't work
try:
ndim, dtype_ = msa.ndim, msa.dtype
except AttributeError:
try:
arr = msa._getArray()
except AttributeError:
raise TypeError('msa must be a character array or an MSA instance')
ndim, dtype_ = arr.ndim, arr.dtype
else:
arr, msa = msa, None
if dtype('|S1') != dtype_:
raise ValueError('msa must be a character array or an MSA instance')
if ndim != 2:
raise ValueError('msa must be a 2D array or an MSA instance')
title = []
cols = None
index = None
if label is not None:
before = arr.shape[1]
LOGGER.timeit('_refine')
try:
upper, lower = label.upper(), label.lower()
except AttributeError:
raise TypeError('label must be a string')
if msa is None:
raise TypeError('msa must be an MSA instance, '
'label cannot be used')
index = msa.getIndex(label)
if index is None:
index = msa.getIndex(upper)
if index is None:
index = msa.getIndex(lower)
chain = None
if index is None and (len(label) == 4 or len(label) == 5):
from prody import parsePDB
try:
structure, header = parsePDB(label[:4], header=True)
except Exception as err:
raise IOError('failed to parse header for {0} ({1})'
.format(label[:4], str(err)))
chid = label[4:].upper()
for poly in header['polymers']:
if chid and poly.chid != chid:
continue
for dbref in poly.dbrefs:
if index is None:
index = msa.getIndex(dbref.idcode)
if index is not None:
LOGGER.info('{0} idcode {1} for {2}{3} '
'is found in chain {3}.'.format(
dbref.database, dbref.idcode,
label[:4], poly.chid, str(msa)))
break
if index is None:
index = msa.getIndex(dbref.accession)
if index is not None:
LOGGER.info('{0} accession {1} for {2}{3} '
'is found in chain {3}.'.format(
dbref.database, dbref.accession,
label[:4], poly.chid, str(msa)))
break
if index is not None:
chain = structure[poly.chid]
if index is None:
raise ValueError('label is not in msa, or msa is not indexed')
try:
len(index)
except TypeError:
pass
else:
raise ValueError('label {0} maps onto multiple sequences, '
'so cannot be used for refinement'.format(label))
title.append('label=' + label)
cols = char.isalpha(arr[index]).nonzero()[0]
arr = arr.take(cols, 1)
LOGGER.report('Label refinement reduced number of columns from {0} to '
'{1} in %.2fs.'.format(before, arr.shape[1]), '_refine')
if chain is not None and not kwargs.get('keep', False):
before = arr.shape[1]
LOGGER.timeit('_refine')
from prody.proteins.compare import importBioPairwise2
from prody.proteins.compare import MATCH_SCORE, MISMATCH_SCORE
from prody.proteins.compare import GAP_PENALTY, GAP_EXT_PENALTY
pw2 = importBioPairwise2()
chseq = chain.getSequence()
algn = pw2.align.localms(arr[index].tostring().upper(), chseq,
MATCH_SCORE, MISMATCH_SCORE,
GAP_PENALTY, GAP_EXT_PENALTY,
one_alignment_only=1)
torf = []
for s, c in zip(*algn[0][:2]):
if s == '-':
continue
elif c != '-':
torf.append(True)
else:
torf.append(False)
torf = array(torf)
tsum = torf.sum()
assert tsum <= before, 'problem in mapping sequence to structure'
if tsum < before:
arr = arr.take(torf.nonzero()[0], 1)
LOGGER.report('Structure refinement reduced number of '
'columns from {0} to {1} in %.2fs.'
.format(before, arr.shape[1]), '_refine')
else:
LOGGER.debug('All residues in the sequence are contained in '
'PDB structure {0}.'.format(label))
from .analysis import calcMSAOccupancy, uniqueSequences
rows = None
if rowocc is not None:
before = arr.shape[0]
LOGGER.timeit('_refine')
try:
rowocc = float(rowocc)
except Exception as err:
raise TypeError('rowocc must be a float ({0})'.format(str(err)))
assert 0. <= rowocc <= 1., 'rowocc must be between 0 and 1'
rows = calcMSAOccupancy(arr, 'row') >= rowocc
if index is not None:
index = rows[:index].sum()
rows = (rows).nonzero()[0]
arr = arr[rows]
title.append('rowocc>=' + str(rowocc))
LOGGER.report('Row occupancy refinement reduced number of rows from '
'{0} to {1} in %.2fs.'.format(before, arr.shape[0]),
'_refine')
if seqid is not None:
before = arr.shape[0]
LOGGER.timeit('_refine')
unique = uniqueSequences(arr, seqid)
if index is not None:
unique[index] = True
unique = unique.nonzero()[0]
arr = arr[unique]
title.append('seqid>=' + str(seqid))
if rows is not None:
rows = rows[unique]
else:
rows = unique
LOGGER.report('Sequence identity refinement reduced number of rows '
'from {0} to {1} in %.2fs.'.format(before, arr.shape[0]),
'_refine')
if colocc is not None:
before = arr.shape[1]
LOGGER.timeit('_refine')
try:
colocc = float(colocc)
except Exception as err:
raise TypeError('colocc must be a float ({0})'.format(str(err)))
assert 0. <= colocc <= 1., 'colocc must be between 0 and 1'
cols = (calcMSAOccupancy(arr, 'col') >= colocc).nonzero()[0]
arr = arr.take(cols, 1)
title.append('colocc>=' + str(colocc))
LOGGER.report('Column occupancy refinement reduced number of columns '
'from {0} to {1} in %.2fs.'.format(before, arr.shape[1]),
'_refine')
if not title:
raise ValueError('label, rowocc, colocc all cannot be None')
# depending on slicing of rows, arr may not have it's own memory
if arr.base is not None:
arr = arr.copy()
if msa is None:
return arr
else:
if rows is None:
from copy import copy
labels = copy(msa._labels)
mapping = copy(msa._mapping)
else:
labels = msa._labels
labels = [labels[i] for i in rows]
mapping = None
return MSA(arr, title=msa.getTitle() + ' refined ({0})'
.format(', '.join(title)), labels=labels, mapping=mapping)
def refineMSA(msa,
index=None,
label=None,
rowocc=None,
seqid=None,
colocc=None,
**kwargs):
"""Refine *msa* by removing sequences (rows) and residues (columns) that
contain gaps.
:arg msa: multiple sequence alignment
:type msa: :class:`.MSA`
:arg index: remove columns that are gaps in the sequence with that index
:type index: int
:arg label: remove columns that are gaps in the sequence matching label,
``msa.getIndex(label)`` must return a sequence index, a PDB identifier
is also acceptable
:type label: str
:arg rowocc: row occupancy, sequences with less occupancy will be
removed after *label* refinement is applied
:type rowocc: float
:arg seqid: keep unique sequences at specified sequence identity level,
unique sequences are identified using :func:`.uniqueSequences`
:type seqid: float
:arg colocc: column occupancy, residue positions with less occupancy
will be removed after other refinements are applied
:type colocc: float
:arg keep: keep columns corresponding to residues not resolved in the PDB
structure, default is **False**, applies when *label* is a PDB
identifier
:arg type: bool
For Pfam MSA data, *label* is UniProt entry name for the protein. You may
also use PDB structure and chain identifiers, e.g. ``'1p38'`` or
``'1p38A'``, for *label* argument and UniProt entry names will be parsed
using :func:`.parsePDBHeader` function (see also :class:`.Polymer` and
:class:`.DBRef`).
The order of refinements are applied in the order of arguments. If *label*
and *unique* is specified, sequence matching *label* will
be kept in the refined :class:`.MSA` although it may be similar to some
other sequence."""
# if msa is a char array, it will be refined but label won't work
try:
ndim, dtype_ = msa.ndim, msa.dtype
except AttributeError:
try:
arr = msa._getArray()
except AttributeError:
raise TypeError('msa must be a character array or an MSA instance')
ndim, dtype_ = arr.ndim, arr.dtype
else:
arr, msa = msa, None
if dtype('|S1') != dtype_:
raise ValueError('msa must be a character array or an MSA instance')
if ndim != 2:
raise ValueError('msa must be a 2D array or an MSA instance')
title = []
cols = None
if index is not None:
before = arr.shape[1]
LOGGER.timeit('_refine')
cols = char.isalpha(arr[index]).nonzero()[0]
arr = arr.take(cols, 1)
title.append('index=' + str(index))
LOGGER.report(
'Index refinement reduced number of columns from {0} to '
'{1} in %.2fs.'.format(before, arr.shape[1]), '_refine')
if label is not None:
if index is not None:
LOGGER.info('An index was provided so the label will be ignored.')
else:
before = arr.shape[1]
LOGGER.timeit('_refine')
try:
upper, lower = label.upper(), label.lower()
except AttributeError:
raise TypeError('label must be a string')
if msa is None:
raise TypeError('msa must be an MSA instance, '
'label cannot be used')
index = msa.getIndex(label)
if index is None:
index = msa.getIndex(upper)
if index is None:
index = msa.getIndex(lower)
chain = None
if index is None and (len(label) == 4 or len(label) == 5):
from prody import parsePDB
try:
structure, header = parsePDB(label[:4], header=True)
except Exception as err:
raise IOError(
'failed to parse header for {0} ({1})'.format(
label[:4], str(err)))
chid = label[4:].upper()
for poly in header['polymers']:
if chid and poly.chid != chid:
continue
for dbref in poly.dbrefs:
if index is None:
index = msa.getIndex(dbref.idcode)
if index is not None:
LOGGER.info('{0} idcode {1} for {2}{3} '
'is found in chain {4}.'.format(
dbref.database, dbref.idcode,
label[:4], poly.chid,
str(msa)))
break
if index is None:
index = msa.getIndex(dbref.accession)
if index is not None:
LOGGER.info('{0} accession {1} for {2}{3} '
'is found in chain {4}.'.format(
dbref.database,
dbref.accession, label[:4],
poly.chid, str(msa)))
break
if index is not None:
chain = structure[poly.chid]
if index is None:
raise ValueError('label is not in msa, or msa is not indexed')
try:
len(index)
except TypeError:
pass
else:
raise ValueError(
'label {0} maps onto multiple sequences, '
'so cannot be used for refinement'.format(label))
title.append('label=' + label)
cols = char.isalpha(arr[index]).nonzero()[0]
arr = arr.take(cols, 1)
LOGGER.report(
'Label refinement reduced number of columns from {0} to '
'{1} in %.2fs.'.format(before, arr.shape[1]), '_refine')
if chain is not None and not kwargs.get('keep', False):
before = arr.shape[1]
LOGGER.timeit('_refine')
from prody.proteins.compare import importBioPairwise2
from prody.proteins.compare import MATCH_SCORE, MISMATCH_SCORE
from prody.proteins.compare import GAP_PENALTY, GAP_EXT_PENALTY
pw2 = importBioPairwise2()
chseq = chain.getSequence()
algn = pw2.align.localms(arr[index].tostring().upper(),
chseq,
MATCH_SCORE,
MISMATCH_SCORE,
GAP_PENALTY,
GAP_EXT_PENALTY,
one_alignment_only=1)
torf = []
for s, c in zip(*algn[0][:2]):
if s == '-':
continue
elif c != '-':
torf.append(True)
else:
torf.append(False)
torf = array(torf)
tsum = torf.sum()
assert tsum <= before, 'problem in mapping sequence to structure'
if tsum < before:
arr = arr.take(torf.nonzero()[0], 1)
LOGGER.report(
'Structure refinement reduced number of '
'columns from {0} to {1} in %.2fs.'.format(
before, arr.shape[1]), '_refine')
else:
LOGGER.debug(
'All residues in the sequence are contained in '
'PDB structure {0}.'.format(label))
from .analysis import calcMSAOccupancy, uniqueSequences
rows = None
if rowocc is not None:
before = arr.shape[0]
LOGGER.timeit('_refine')
try:
rowocc = float(rowocc)
except Exception as err:
raise TypeError('rowocc must be a float ({0})'.format(str(err)))
assert 0. <= rowocc <= 1., 'rowocc must be between 0 and 1'
rows = calcMSAOccupancy(arr, 'row') >= rowocc
if index is not None:
index = rows[:index].sum()
rows = (rows).nonzero()[0]
arr = arr[rows]
title.append('rowocc>=' + str(rowocc))
LOGGER.report(
'Row occupancy refinement reduced number of rows from '
'{0} to {1} in %.2fs.'.format(before, arr.shape[0]), '_refine')
if seqid is not None:
before = arr.shape[0]
LOGGER.timeit('_refine')
unique = uniqueSequences(arr, seqid)
if index is not None:
unique[index] = True
unique = unique.nonzero()[0]
arr = arr[unique]
title.append('seqid>=' + str(seqid))
if rows is not None:
rows = rows[unique]
else:
rows = unique
LOGGER.report(
'Sequence identity refinement reduced number of rows '
'from {0} to {1} in %.2fs.'.format(before, arr.shape[0]),
'_refine')
if colocc is not None:
before = arr.shape[1]
LOGGER.timeit('_refine')
try:
colocc = float(colocc)
except Exception as err:
raise TypeError('colocc must be a float ({0})'.format(str(err)))
assert 0. <= colocc <= 1., 'colocc must be between 0 and 1'
cols = (calcMSAOccupancy(arr, 'col') >= colocc).nonzero()[0]
arr = arr.take(cols, 1)
title.append('colocc>=' + str(colocc))
LOGGER.report(
'Column occupancy refinement reduced number of columns '
'from {0} to {1} in %.2fs.'.format(before, arr.shape[1]),
'_refine')
if not title:
raise ValueError(
'label, index, seqid, rowocc, colocc all cannot be None')
# depending on slicing of rows, arr may not have it's own memory
if arr.base is not None:
arr = arr.copy()
if msa is None:
return arr
else:
if rows is None:
from copy import copy
labels = copy(msa._labels)
else:
labels = msa._labels
labels = [labels[i] for i in rows]
return MSA(arr,
title=msa.getTitle() +
' refined ({0})'.format(', '.join(title)),
labels=labels)