def muscle_seqs(seqs,
add_seq_names=False,
out_filename=None,
input_handler=None,
params={},
WorkingDir=None,
SuppressStderr=None,
SuppressStdout=None):
"""Muscle align list of sequences.
seqs: a list of sequences as strings or objects, you must set add_seq_names=True
or sequences in a multiline string, as read() from a fasta file
or sequences in a list of lines, as readlines() from a fasta file
or a fasta seq filename.
== for eg, testcode for guessing
#guess_input_handler should correctly identify input
gih = guess_input_handler
self.assertEqual(gih('abc.txt'), '_input_as_string')
self.assertEqual(gih('>ab\nTCAG'), '_input_as_multiline_string')
self.assertEqual(gih(['ACC','TGA'], True), '_input_as_seqs')
self.assertEqual(gih(['>a','ACC','>b','TGA']), '_input_as_lines')
== docstring for blast_seqs, apply to muscle_seqs ==
seqs: either file name or list of sequence objects or list of strings or
single multiline string containing sequences.
WARNING: DECISION RULES FOR INPUT HANDLING HAVE CHANGED. Decision rules
for data are as follows. If it's s list, treat as lines, unless
add_seq_names is true (in which case treat as list of seqs). If it's a
string, test whether it has newlines. If it doesn't have newlines, assume
it's a filename. If it does have newlines, it can't be a filename, so
assume it's a multiline string containing sequences.
If you want to skip the detection and force a specific type of input
handler, use input_handler='your_favorite_handler'.
add_seq_names: boolean. if True, sequence names are inserted in the list
of sequences. if False, it assumes seqs is a list of lines of some
proper format that the program can handle
Addl docs coming soon
"""
if out_filename:
params["-out"] = out_filename
#else:
# params["-out"] = get_tmp_filename(WorkingDir)
ih = input_handler or guess_input_handler(seqs, add_seq_names)
muscle_app = Muscle(
params=params,
InputHandler=ih,
WorkingDir=WorkingDir,
SuppressStderr=SuppressStderr,
SuppressStdout=SuppressStdout)
return muscle_app(seqs)
def muscle_seqs(seqs,
add_seq_names=False,
out_filename=None,
input_handler=None,
params={},
WorkingDir=None,
SuppressStderr=None,
SuppressStdout=None):
"""Muscle align list of sequences.
seqs: a list of sequences as strings or objects, you must set add_seq_names=True
or sequences in a multiline string, as read() from a fasta file
or sequences in a list of lines, as readlines() from a fasta file
or a fasta seq filename.
== for eg, testcode for guessing
#guess_input_handler should correctly identify input
gih = guess_input_handler
self.assertEqual(gih('abc.txt'), '_input_as_string')
self.assertEqual(gih('>ab\nTCAG'), '_input_as_multiline_string')
self.assertEqual(gih(['ACC','TGA'], True), '_input_as_seqs')
self.assertEqual(gih(['>a','ACC','>b','TGA']), '_input_as_lines')
== docstring for blast_seqs, apply to muscle_seqs ==
seqs: either file name or list of sequence objects or list of strings or
single multiline string containing sequences.
WARNING: DECISION RULES FOR INPUT HANDLING HAVE CHANGED. Decision rules
for data are as follows. If it's s list, treat as lines, unless
add_seq_names is true (in which case treat as list of seqs). If it's a
string, test whether it has newlines. If it doesn't have newlines, assume
it's a filename. If it does have newlines, it can't be a filename, so
assume it's a multiline string containing sequences.
If you want to skip the detection and force a specific type of input
handler, use input_handler='your_favorite_handler'.
add_seq_names: boolean. if True, sequence names are inserted in the list
of sequences. if False, it assumes seqs is a list of lines of some
proper format that the program can handle
Addl docs coming soon
"""
if out_filename:
params["-out"] = out_filename
#else:
# params["-out"] = get_tmp_filename(WorkingDir)
ih = input_handler or guess_input_handler(seqs, add_seq_names)
muscle_app = Muscle(
params=params,
InputHandler=ih,
WorkingDir=WorkingDir,
SuppressStderr=SuppressStderr,
SuppressStdout=SuppressStdout)
return muscle_app(seqs)
def blast_seqs(seqs,
blast_constructor,
blast_db=None,
blast_mat_root=None,
params={},
add_seq_names=True,
out_filename=None,
WorkingDir=None,
SuppressStderr=None,
SuppressStdout=None,
input_handler=None,
HALT_EXEC=False
):
"""Blast list of sequences.
seqs: either file name or list of sequence objects or list of strings or
single multiline string containing sequences.
WARNING: DECISION RULES FOR INPUT HANDLING HAVE CHANGED. Decision rules
for data are as follows. If it's s list, treat as lines, unless
add_seq_names is true (in which case treat as list of seqs). If it's a
string, test whether it has newlines. If it doesn't have newlines, assume
it's a filename. If it does have newlines, it can't be a filename, so
assume it's a multiline string containing sequences.
If you want to skip the detection and force a specific type of input
handler, use input_handler='your_favorite_handler'.
add_seq_names: boolean. if True, sequence names are inserted in the list
of sequences. if False, it assumes seqs is a list of lines of some
proper format that the program can handle
"""
# set num keep
if blast_db:
params["-d"] = blast_db
if out_filename:
params["-o"] = out_filename
ih = input_handler or guess_input_handler(seqs, add_seq_names)
blast_app = blast_constructor(
params=params,
blast_mat_root=blast_mat_root,
InputHandler=ih,
WorkingDir=WorkingDir,
SuppressStderr=SuppressStderr,
SuppressStdout=SuppressStdout,
HALT_EXEC=HALT_EXEC)
return blast_app(seqs)
def blast_seqs(seqs,
blast_constructor,
blast_db=None,
blast_mat_root=None,
params={},
add_seq_names=True,
out_filename=None,
WorkingDir=None,
SuppressStderr=None,
SuppressStdout=None,
input_handler=None,
HALT_EXEC=False):
"""Blast list of sequences.
seqs: either file name or list of sequence objects or list of strings or
single multiline string containing sequences.
WARNING: DECISION RULES FOR INPUT HANDLING HAVE CHANGED. Decision rules
for data are as follows. If it's s list, treat as lines, unless
add_seq_names is true (in which case treat as list of seqs). If it's a
string, test whether it has newlines. If it doesn't have newlines, assume
it's a filename. If it does have newlines, it can't be a filename, so
assume it's a multiline string containing sequences.
If you want to skip the detection and force a specific type of input
handler, use input_handler='your_favorite_handler'.
add_seq_names: boolean. if True, sequence names are inserted in the list
of sequences. if False, it assumes seqs is a list of lines of some
proper format that the program can handle
"""
# set num keep
if blast_db:
params["-d"] = blast_db
if out_filename:
params["-o"] = out_filename
ih = input_handler or guess_input_handler(seqs, add_seq_names)
blast_app = blast_constructor(params=params,
blast_mat_root=blast_mat_root,
InputHandler=ih,
WorkingDir=WorkingDir,
SuppressStderr=SuppressStderr,
SuppressStdout=SuppressStdout,
HALT_EXEC=HALT_EXEC)
return blast_app(seqs)
def psiblast_n_neighbors(seqs,
n=100,
blast_db=None,
core_threshold=1e-50,
extra_threshold=1e-10,
lower_threshold=1e-6,
step=100,
method="two-step",
blast_mat_root=None,
params={},
add_seq_names=False,
WorkingDir=None,
SuppressStderr=None,
SuppressStdout=None,
input_handler=None,
scorer=3, #shotgun with 3 hits needed to keep
second_db=None
):
"""PsiBlasts sequences, stopping when n neighbors are reached.
core_threshold: threshold for the core profile (default: 1e-50)
extra_threshold: threshold for pulling in additional seqs (default:1e-10)
lower_threshold: threshold for seqs in final round (default:1e-6)
seqs: either file name or list of sequence objects or list of strings or
single multiline string containing sequences.
If you want to skip the detection and force a specific type of input
handler, use input_handler='your_favorite_handler'.
add_seq_names: boolean. if True, sequence names are inserted in the list
of sequences. if False, it assumes seqs is a list of lines of some
proper format that the program can handle
"""
if blast_db:
params["-d"] = blast_db
ih = input_handler or guess_input_handler(seqs, add_seq_names)
recs = seqs_to_stream(seqs, ih) #checkpointing can only handle one seq...
#set up the parameters for the core and additional runs
max_iterations = params['-j']
params['-j'] = 2 #won't checkpoint with single iteration
app = PsiBlast(params=params,
blast_mat_root=blast_mat_root,
InputHandler='_input_as_lines',
WorkingDir=WorkingDir,
SuppressStderr=SuppressStderr,
SuppressStdout=SuppressStdout,
)
result = {}
for seq in recs:
query_id = seq[0][1:].split(None,1)[0]
if method == "two-step":
result[query_id] = ids_from_seq_two_step(seq, n, max_iterations, \
app, core_threshold, extra_threshold, lower_threshold, second_db)
elif method == "lower_threshold":
result[query_id] = ids_from_seq_lower_threshold(seq, n, \
max_iterations, app, core_threshold, lower_threshold, step)
elif method == "iterative":
result[query_id] = ids_from_seqs_iterative(seq, app, \
QMEPsiBlast9, scorer, params['-j'], n)
else:
raise TypeError("Got unknown method %s" % method)
params['-j'] = max_iterations
return result
def psiblast_n_neighbors(
seqs,
n=100,
blast_db=None,
core_threshold=1e-50,
extra_threshold=1e-10,
lower_threshold=1e-6,
step=100,
method="two-step",
blast_mat_root=None,
params={},
add_seq_names=False,
WorkingDir=None,
SuppressStderr=None,
SuppressStdout=None,
input_handler=None,
scorer=3, #shotgun with 3 hits needed to keep
second_db=None):
"""PsiBlasts sequences, stopping when n neighbors are reached.
core_threshold: threshold for the core profile (default: 1e-50)
extra_threshold: threshold for pulling in additional seqs (default:1e-10)
lower_threshold: threshold for seqs in final round (default:1e-6)
seqs: either file name or list of sequence objects or list of strings or
single multiline string containing sequences.
If you want to skip the detection and force a specific type of input
handler, use input_handler='your_favorite_handler'.
add_seq_names: boolean. if True, sequence names are inserted in the list
of sequences. if False, it assumes seqs is a list of lines of some
proper format that the program can handle
"""
if blast_db:
params["-d"] = blast_db
ih = input_handler or guess_input_handler(seqs, add_seq_names)
recs = seqs_to_stream(seqs, ih) #checkpointing can only handle one seq...
#set up the parameters for the core and additional runs
max_iterations = params['-j']
params['-j'] = 2 #won't checkpoint with single iteration
app = PsiBlast(
params=params,
blast_mat_root=blast_mat_root,
InputHandler='_input_as_lines',
WorkingDir=WorkingDir,
SuppressStderr=SuppressStderr,
SuppressStdout=SuppressStdout,
)
result = {}
for seq in recs:
query_id = seq[0][1:].split(None, 1)[0]
if method == "two-step":
result[query_id] = ids_from_seq_two_step(seq, n, max_iterations, \
app, core_threshold, extra_threshold, lower_threshold, second_db)
elif method == "lower_threshold":
result[query_id] = ids_from_seq_lower_threshold(seq, n, \
max_iterations, app, core_threshold, lower_threshold, step)
elif method == "iterative":
result[query_id] = ids_from_seqs_iterative(seq, app, \
QMEPsiBlast9, scorer, params['-j'], n)
else:
raise TypeError, "Got unknown method %s" % method
params['-j'] = max_iterations
return result
