def run_modeller(knowns = ('2B4C', '4NCO'), seq = 'CAP45',
num_mod = 1):
seq_dir = '../sequence_files/'
pdb_dir = '../pdb_files/'
j = job()
j.append(local_slave())
j.append(local_slave())
j.append(local_slave())
j.append(local_slave())
j.append(local_slave())
# Modeller environment
env = environ()
# Output
log.none()
#input dir
env.io.atom_files_directory = [pdb_dir]
#input file
ali_file = seq_dir + 'align' + seq
if isinstance(knowns, list):
for i in range(len(knowns)):
ali_file += '_' + knowns[i]
ali_file += '.ali'
else:
ali_file += '_' + str(knowns) + '.ali'
mod = automodel(env,
alnfile = ali_file,
knowns = knowns,
sequence = seq,
assess_methods = (assess.DOPE))
mod.starting_model = 1
mod.ending_model = num_mod
mod.use_parallel_job(j)
mod.make()
# Get list of all built models
ok_models = [x for x in mod.outputs if x['failure'] is None]
# Rank the models by DOPE score
ok_models.sort(key = lambda mod: mod['DOPE score'])
# Get top model
m = ok_models[0]
print("The best model {} with a dope of {}".format(m['name'], m['DOPE score']))
def modeller_automodel(self, query: SeqRecord, results: Path,
num_align: int, atom_files_dir: Path):
from modeller import environ
from modeller.automodel import automodel
for model_index, r in enumerate(
np.load(results, allow_pickle=True)[:num_align]):
try:
aln = AlignIO.read(StringIO(r[-2][0]), 'clustal')
except:
logging.error(
f'Failed to parse alignment: {r[0]} -> {r[2]} -> {r[4]} -> {r[6]}'
)
continue
assert query.id == aln[0].id and aln[-1].id == r[-3]
q_rec, t_rec = self._remove_gaps(aln[0], aln[-1])
try:
t_rec = self._remove_missing_res(
t_rec, (atom_files_dir / aln[-1].id[2:4] /
f'{aln[-1].id}.ent').resolve().as_posix())
except FileNotFoundError as e:
logging.exception(e)
continue
q_rec.name, t_rec.name = '', ''
q_rec.description = f'sequence:{q_rec.id}::::::::'
t_rec.description = f'structureX:{t_rec.id}::{t_rec.id[-2].upper()}::{t_rec.id[-2].upper()}::::'
aln = MultipleSeqAlignment([q_rec, t_rec])
out_d = results.resolve().parent
if (out_d / f'{aln[0].id}_{model_index+1}.pdb').exists():
continue
cwd = os.getcwd()
with tempfile.TemporaryDirectory() as tmpdir:
try:
os.chdir(tmpdir)
AlignIO.write(aln, 'aln.pir', 'pir')
env = environ()
env.io.atom_files_directory = [
(atom_files_dir / aln[1].id[2:4]).resolve().as_posix()
]
mod = automodel(env,
'aln.pir',
knowns=[aln[1].id],
sequence=aln[0].id)
mod.make()
shutil.copy(
list(Path().glob('*.B*.pdb'))[0],
out_d / f'{aln[0].id}_{model_index+1}.pdb')
except Exception as e:
logging.error(
f'knowns=[{aln[1].id}], sequence={aln[0].id}')
logging.exception(e)
finally:
os.chdir(cwd)
def _automodel_run(self, alin_file, knowns):
amdl = automodel(self.env,
alnfile=alin_file,
knowns=knowns,
sequence='target',
assess_methods=(assess.DOPE, assess.GA341))
amdl.starting_model = 1
amdl.ending_model = 1
orig_dir = os.getcwd()
os.chdir(self.tmpdir)
amdl.make()
os.chdir(orig_dir)
return amdl.outputs[0]
def create_models(alnfile,
knownid,
sequenceid,
pdbfile,
model_number=5,
dir=''):
if dir:
cdir = os.getcwd()
ndir = cdir + '/' + dir
os.mkdir(ndir)
os.chdir(ndir)
sp.check_call("cp ../" + pdbfile + ' ' + pdbfile, shell=True)
try:
from modeller import environ
from modeller.automodel import automodel
except ImportError:
print(
'Make Sure Python Modeller is installed. Double check License Key is in Modeller config.py file'
)
sys.exit()
env = environ()
if dir:
a = automodel(env,
alnfile='../' + alnfile,
knowns=knownid,
sequence=sequenceid)
else:
a = automodel(env,
alnfile=alnfile,
knowns=knownid,
sequence=sequenceid)
a.starting_model = 1
a.ending_model = model_number
a.make()
if dir:
os.chdir(cdir)
def model3D(fic, ALLPDB, pdb_extension='.cif'):
if _MODELLER_MESSAGE != "":
raise ImportError(_MODELLER_MESSAGE)
seqs = readFastaMul(fic)
if len(seqs) < 2:
raise Exception("There aren't template sequences in %s." % fic)
seq = seqs[0][0].split('\n')[0]
seq = seq.split(';')[1]
modeller.log.verbose() # request verbose output
env = modeller.environ() # create a new MODELLER environment to build ...
# ... this model in
env.io.atom_files_directory = ['.'] # ['.', ALLPDB]
knowns = []
for i in range(1, len(seqs)):
tmp = seqs[i][0].split('\n')[0]
tmp = tmp.split(';')[1]
knowns.append(tmp)
kn = tmp.split('_')[0].lower() # upper ?
base_name = kn + pdb_extension
nam = base_name # + '.gz'
if not os.path.isfile(kn + pdb_extension):
# shutil.copyfile(ALLPDB + nam, "./" + nam)
shutil.copy2(os.path.join(ALLPDB, nam), nam)
# # os.system("gunzip ./" + nam)
# with gzip.open(nam, 'rb') as f_in:
# with open(base_name, 'wb') as f_out:
# shutil.copyfileobj(f_in, f_out)
# splitChainsPDB('pdb' + kn + pdb_extension, kn, 'pdb')
knowns = tuple(knowns)
a = automodel.automodel(env, alnfile=fic, knowns=knowns, sequence=seq)
a.max_molpdf = 1e12
a.starting_model = 1 # index of the first model
a.ending_model = 1 # index of the last model
# (determines how many models to calculate)
cdir = os.getcwd()
print("cdir = ", cdir)
a.make() # do the actual homology modeling
return 1
def run(self, seqId):
assert (self.check(seqId))
from modeller import log, environ
from modeller.automodel import automodel
#log.verbose()
#log.none()
log.level(output=0, errors=0, notes=0, warnings=0, memory=0)
env = environ()
env.io.atom_files_directory = [self.template[0]]
a = automodel(env,
alnfile=self.pir,
knowns=self.template[1],
sequence=seqId)
a.auto_align()
a.make()
fn = '{}.B99990001.pdb'.format(seqId)
assert (os.path.isfile(fn))
return os.path.abspath(fn)
def modeller_automodel(self, query: SeqRecord, results: Path,
num_align: int, atom_files_dir: Path):
from modeller import environ, log
from modeller.automodel import automodel
raw_df = pandas.read_csv(
'data/delta_new_hits_thresh10_xdug10_xdg10_xdgf10.csv')
for row in tqdm(raw_df.itertuples(), total=raw_df.shape[0]):
try:
aln = SearchIO.read(StringIO(row.XML), 'blast-xml').hsps[0].aln
except IndexError:
continue
assert aln[0].id == row.Query and aln[1].id == row.Hit
q_rec, t_rec = self._remove_gaps(aln[0], aln[-1])
q_rec.name = ''
t_rec.name = ''
q_rec.description = f'sequence:{row.Query}::::::::'
t_rec.description = f'structureX:{row.Hit}::{row.Hit[-2].upper()}::{row.Hit[-2].upper()}::::'
aln = MultipleSeqAlignment([q_rec, t_rec])
out_d = results.resolve()
cwd = os.getcwd()
with tempfile.TemporaryDirectory() as tmpdir:
try:
os.chdir(tmpdir)
AlignIO.write(aln, 'aln.pir', 'pir')
log.none()
env = environ()
env.io.atom_files_directory = [
(atom_files_dir / aln[1].id[2:4]).resolve().as_posix()
]
mod = automodel(env,
'aln.pir',
knowns=[aln[1].id],
sequence=aln[0].id)
mod.make()
shutil.copy(
list(Path().glob('*.B*.pdb'))[0],
out_d / f'{aln[0].id}_{aln[1].id}.pdb')
except:
pass
finally:
os.chdir(cwd)
def main(args):
mod.log.verbose()
env = mod.environ(rand_seed=args.seed)
env.io.atom_files_directory = [".", args.dir, "../" + args.dir]
seq = args.target.replace(".ali", "")
alnfile = os.path.join(args.dir, seq + "-" + args.template.replace(".pdb", ".ali"))
# in order to use the soap assess method, refer to https://salilab.org/SOAP/ and
# download the SOAP-Protein library file. Put this file in
# your-installation-path/lib/modeller-9.25/modlib/
# otherwise, just comment the respective line
a = automodel.automodel(
env,
alnfile=alnfile,
knowns=args.template.replace(".pdb", ""),
sequence=seq,
assess_methods=(
automodel.assess.DOPE,
automodel.assess.GA341,
# soap_protein_od.Scorer(),
),
)
a.starting_model, a.ending_model = 1, args.num_models
a.make()
# get list of all successfully built models from a.outputs
models = [m for m in a.outputs if m["failure"] is None]
key = "DOPE score"
models.sort(key=lambda a: a[key])
# print and return top model DOPE score
top_model = models[0]
print("Top model: %s (DOPE score %.3f)" % (top_model["name"], top_model[key]))
return top_model["name"]
def process(req, rep):
'''Processes a single request to the server, storing the result in `rep`'''
from modeller.automodel.assess import DOPE, GA341
from modeller.automodel import automodel
logger = logging.getLogger('modeller_server')
logger.info('Processing job=%s, recipient=%s, alignments=%d' % (req.identifier, req.recipient, len(req.alignments)))
# In order to prevent filename collisions, independent runs of modeller
# are executed in separate directories
curr_dir = os.getcwd()
work_dir = tempfile.mkdtemp()
os.chdir(work_dir)
# Populate required fields in the response
rep.recipient = req.recipient
rep.identifier = req.identifier
# N * M (model, alignment, score) tuples, where N = #alignments and M = #models per alignment
candidates = []
for alignment in req.alignments:
query_id = 'query'
templ_id = str(alignment.templ_pdb + alignment.templ_chain)
# Write template structure
templ_file = templ_id + '.pdb'
with open(templ_file, 'w') as file:
file.write('%s\n' % alignment.templ_structure)
# Write alignment
alignment_file = templ_id + '.ali'
with open(alignment_file, 'w') as file:
params = { 'query_id' : query_id,
'query_start' : alignment.query_start,
'query_stop' : alignment.query_stop,
'query_align' : alignment.query_align }
query_line = query_alignment.safe_substitute(params)
file.write('%s\n' % query_line)
params = { 'templ_id' : templ_id,
'templ_pdb' : templ_file,
'templ_chain' : alignment.templ_chain,
'templ_start' : alignment.templ_start,
'templ_stop' : alignment.templ_stop,
'templ_align' : alignment.templ_align }
templ_line = templ_alignment.safe_substitute(params)
file.write('%s\n' % templ_line)
# Run modeler
modeller.log.verbose()
env = modeller.environ()
env.io.atom_files_directory = ['.']
am = automodel(env,
alnfile = alignment_file,
knowns = templ_id,
sequence = query_id,
assess_methods = (DOPE, GA341))
am.starting_model = 1
am.ending_model = FLAGS.models_per_alignment
am.make()
# Rank successful predictions by DOPE score
models = [x for x in am.outputs if x['failure'] is None]
models.sort(key = lambda x: x['DOPE score'])
logger.info('Produced %d models for alignment %s' % (len(models), templ_id))
for model in models:
with open(model['name']) as file:
coords = ''
for line in file:
coords += line
entry = (coords, alignment, model['DOPE score'])
candidates.append(entry)
# Sort all N * M candidate models in increasing order of DOPE score, returning the top K
candidates.sort(key = operator.itemgetter(-1))
for (i, entry) in enumerate(candidates):
coords, alignment, score = entry
selection = rep.selected.add()
selection.rank = i + 1
# Append alignment information to bottom of PDB file
selection.model = coords
selection.model += 'Source: %s\n' % alignment.method
selection.model += 'Template: %s\n' % templ_id
selection.model += 'Query alignment: %s\n' % alignment.query_align
selection.model += 'Templ alignment: %s\n' % alignment.templ_align
# Message types cannot be assigned directory (e.g. x.field = field).
# For additional details, read the "Singular Message Fields" section in:
# https://developers.google.com/protocol-buffers/docs/reference/python-generated#fields
selection.alignment.ParseFromString(alignment.SerializeToString())
if (selection.rank == FLAGS.max_models_to_return):
break
os.chdir(curr_dir)
shutil.rmtree(work_dir)
logger.info('Completed job=%s, recipient=%s' % (req.identifier, req.recipient))
def peptide_rebuild_modeller(name, selection='all', hetatm=0, sequence=None,
nmodels=1, hydro=0, quiet=1, *, _self=cmd):
'''
DESCRIPTION
Remodel the given selection using modeller. This is useful for example to
build incomplete sidechains. More complicated modelling tasks are not
the intention of this simple interface.
Side effects: Alters "type" property for MSE residues in selection
(workaround for bug #3512313).
USAGE
peptide_rebuild_modeller name [, selection [, hetatm [, sequence ]]]
ARGUMENTS
name = string: new object name
selection = string: atom selection
hetatm = 0/1: read and model HETATMs (ligands) {default: 0}
sequence = string: if provided, use this sequence instead of the
template sequence {default: None}
nmodels = int: number of models (states) to generate {default: 1}
'''
import modeller
from modeller.automodel import automodel, allhmodel
import tempfile, shutil, os
_assert_package_import()
from .editing import update_identifiers
nmodels, hetatm, quiet = int(nmodels), int(hetatm), int(quiet)
if int(hydro):
automodel = allhmodel
tempdir = tempfile.mkdtemp()
pdbfile = os.path.join(tempdir, 'template.pdb')
alnfile = os.path.join(tempdir, 'aln.pir')
cwd = os.getcwd()
os.chdir(tempdir)
if not quiet:
print(' Notice: PWD=%s' % (tempdir))
try:
modeller.log.none()
env = modeller.environ()
env.io.hetatm = hetatm
# prevent PyMOL to put TER records before MSE residues (bug #3512313)
_self.alter('(%s) and polymer' % (selection), 'type="ATOM"')
_self.save(pdbfile, selection)
mdl = modeller.model(env, file=pdbfile)
aln = modeller.alignment(env)
aln.append_model(mdl, align_codes='foo', atom_files=pdbfile)
# get sequence from non-present atoms
if not sequence and _self.count_atoms('(%s) & !present' % (selection)):
sequence = get_seq(selection)
if sequence:
aln.append_sequence(sequence)
aln[-1].code = 'bar'
aln.malign()
aln.write(alnfile)
a = automodel(env, alnfile=alnfile, sequence=aln[-1].code,
knowns=[s.code for s in aln if s.prottyp.startswith('structure')])
a.max_ca_ca_distance = 30.0
if nmodels > 1:
a.ending_model = nmodels
from multiprocessing import cpu_count
ncpu = min(cpu_count(), nmodels)
if ncpu > 1:
from modeller import parallel
job = parallel.job(parallel.local_slave()
for _ in range(ncpu))
a.use_parallel_job(job)
a.make()
for output in a.outputs:
_self.load(output['name'], name, quiet=quiet)
finally:
os.chdir(cwd)
shutil.rmtree(tempdir)
_self.align(name, selection, cycles=0)
if not sequence:
update_identifiers(name, selection, _self=_self)
if not quiet:
print(' peptide_rebuild_modeller: done')
def run_modeller(structure, alignment, temp_dir: Union[str, Path, Callable]):
"""Run Modeller to create a homology model.
Args:
structure: Structure of the template protein.
alignment_file: Alignment of the target sequence(s) to chain(s) of the template structure.
temp_dir: Location to use for storing Modeller temporary files and output.
Returns:
results: A dictionary of model properties. Of particular interest are the followng:
`name`: The name of the generated PDB structure.
`Normalized DOPE score`: DOPE score that should be comparable between structures.
`GA341 score`: GA341 score that should be comparable between structures.
"""
import modeller
from modeller.automodel import assess, automodel, autosched
if isinstance(structure, (str, Path)):
structure = PDB.load(structure)
if callable(temp_dir):
temp_dir = Path(temp_dir())
else:
temp_dir = Path(temp_dir)
assert len(alignment) == 2
target_id = alignment[0].id
template_id = alignment[1].id
PDB.save(structure, temp_dir.joinpath(f"{template_id}.pdb"))
alignment_file = temp_dir.joinpath(f"{template_id}-{target_id}.aln")
structure_tools.write_pir_alignment(alignment, alignment_file)
# Don't display log messages
modeller.log.none()
# Create a new MODELLER environment
env = modeller.environ()
# Directories for input atom files
env.io.atom_files_directory = [str(temp_dir)]
env.schedule_scale = modeller.physical.values(default=1.0, soft_sphere=0.7)
# Selected atoms do not feel the neighborhood
# env.edat.nonbonded_sel_atoms = 2
env.io.hetatm = True # read in HETATM records from template PDBs
env.io.water = True # read in WATER records (including waters marked as HETATMs)
a = automodel(
env,
# alignment filename
alnfile=str(alignment_file),
# codes of the templates
knowns=(str(template_id)),
# code of the target
sequence=str(target_id),
# wich method for validation should be calculated
assess_methods=(assess.DOPE, assess.normalized_dope, assess.GA341),
)
a.starting_model = 1 # index of the first model
a.ending_model = 1 # index of the last model
# Very thorough VTFM optimization:
a.library_schedule = autosched.slow
a.max_var_iterations = 300
# Thorough MD optimization:
# a.md_level = refine.slow
a.md_level = None
# a.repeat_optimization = 2
# Stop if the objective function is higher than this value
a.max_molpdf = 2e6
with py_tools.log_print_statements(logger), system_tools.switch_paths(temp_dir):
a.make()
assert len(a.outputs) == 1
return a.outputs[0]
def __run_modeller(self, alignFile, loopRefinement):
""".
Parameters
----------
alignFile : string
File containing the input data
result : list
The successfully calculated models are stored in this list
loopRefinement : boolean
If `True`, perform loop refinements
Returns
-------
list
Successfully calculated models
"""
log.none() # instructs Modeller to display no log output.
env = environ() # create a new MODELLER environment to build this model in
# Directories for input atom files
env.io.atom_files_directory = [str(self.filePath.rstrip('/')), ]
env.schedule_scale = physical.values(default=1.0, soft_sphere=0.7)
# Selected atoms do not feel the neighborhood
# env.edat.nonbonded_sel_atoms = 2
env.io.hetatm = True # read in HETATM records from template PDBs
env.io.water = True # read in WATER records (including waters marked as HETATMs)
logger.debug(
'Performing loop refinement in addition to regular modelling: {}'
.format(loopRefinement)
)
if not loopRefinement:
a = automodel(
env,
# alignment filename
alnfile=str(alignFile),
# codes of the templates
knowns=(str(self.templateID)),
# code of the target
sequence=str(self.seqID),
# wich method for validation should be calculated
assess_methods=(assess.DOPE, assess.normalized_dope)
)
else:
a = dope_loopmodel(
env,
# alignment filename
alnfile=str(alignFile),
# codes of the templates
knowns=(str(self.templateID)),
# code of the target
sequence=str(self.seqID),
# wich method for validation should be calculated
assess_methods=(assess.DOPE, assess.normalized_dope),
loop_assess_methods=(assess.DOPE, assess.normalized_dope)
)
# index of the first loop model
a.loop.starting_model = self.loopStart
# index of the last loop model
a.loop.ending_model = self.loopEnd
# loop refinement method; this yields
a.loop.md_level = refine.slow
a.starting_model = self.start # index of the first model
a.ending_model = self.end # index of the last model
# Very thorough VTFM optimization:
a.library_schedule = autosched.slow
a.max_var_iterations = 300
# Thorough MD optimization:
# a.md_level = refine.slow
a.md_level = None
# Repeat the whole cycle 2 times and do not stop unless obj.func. > 1E6
# a.repeat_optimization = 2
a.max_molpdf = 2e5
# with helper.print_heartbeats(): # use 'long_wait' in .travis.yml
with helper.log_print_statements(logger):
a.make() # do the actual homology modeling
# The output produced by modeller is stored in a.loop.outputs or a.outputs
# it is a dictionary
# Check for each model if it was successfully calculated, i.e.
# for each "normal" model and each loop model and append the
# assessment score to a list which is used to return the best model
result = []
loop = False
failures = []
# Add the normal output
for i in range(len(a.outputs)):
if not a.outputs[i]['failure']:
model_filename = a.outputs[i]['name']
model_dope_score = a.outputs[i]['Normalized DOPE score']
logger.debug(
'Success! model_filename: {}, model_dope_score: {}'
.format(model_filename, model_dope_score))
result.append((model_filename, model_dope_score))
else:
failure = a.outputs[i]['failure']
logger.debug('Failure! {}'.format(failure))
failures.append(a.outputs[i]['failure'])
# Add the loop refinement output
if loopRefinement:
logger.debug('Modeller loop outputs:')
for i in range(len(a.loop.outputs)):
if not a.loop.outputs[i]['failure']:
model_filename = a.loop.outputs[i]['name']
model_dope_score = a.loop.outputs[i]['Normalized DOPE score']
logger.debug(
'Success! model_filename: {}, model_dope_score: {}'
.format(model_filename, model_dope_score))
result.append((model_filename, model_dope_score))
loop = True
else:
failure = a.loop.outputs[i]['failure']
logger.debug('Failure! {}'.format(failure))
failures.append(failure)
# Return the successfully calculated models and a loop flag indicating
# whether the returned models are loop refined or not
return result, loop, failures
def peptide_rebuild_modeller(name, selection='all', hetatm=0, sequence=None,
nmodels=1, hydro=0, quiet=1):
'''
DESCRIPTION
Remodel the given selection using modeller. This is useful for example to
build incomplete sidechains. More complicated modelling tasks are not
the intention of this simple interface.
Side effects: Alters "type" property for MSE residues in selection
(workaround for bug #3512313).
USAGE
peptide_rebuild_modeller name [, selection [, hetatm [, sequence ]]]
ARGUMENTS
name = string: new object name
selection = string: atom selection
hetatm = 0/1: read and model HETATMs (ligands) {default: 0}
sequence = string: if provided, use this sequence instead of the
template sequence {default: None}
nmodels = int: number of models (states) to generate {default: 1}
'''
try:
import modeller
from modeller.automodel import automodel, allhmodel
except ImportError:
print(' Error: failed to import "modeller"')
raise CmdException
import tempfile, shutil, os
from .editing import update_identifiers
nmodels, hetatm, quiet = int(nmodels), int(hetatm), int(quiet)
if int(hydro):
automodel = allhmodel
tempdir = tempfile.mkdtemp()
pdbfile = os.path.join(tempdir, 'template.pdb')
alnfile = os.path.join(tempdir, 'aln.pir')
cwd = os.getcwd()
os.chdir(tempdir)
if not quiet:
print(' Notice: PWD=%s' % (tempdir))
try:
modeller.log.none()
env = modeller.environ()
env.io.hetatm = hetatm
# prevent PyMOL to put TER records before MSE residues (bug #3512313)
cmd.alter('(%s) and polymer' % (selection), 'type="ATOM"')
cmd.save(pdbfile, selection)
mdl = modeller.model(env, file=pdbfile)
aln = modeller.alignment(env)
aln.append_model(mdl, align_codes='foo', atom_files=pdbfile)
if sequence:
aln.append_sequence(sequence)
aln[-1].code = 'bar'
aln.malign()
aln.write(alnfile)
a = automodel(env, alnfile=alnfile, sequence=aln[-1].code,
knowns=[s.code for s in aln if s.prottyp.startswith('structure')])
a.max_ca_ca_distance = 30.0
if nmodels > 1:
a.ending_model = nmodels
from multiprocessing import cpu_count
ncpu = min(cpu_count(), nmodels)
if ncpu > 1:
from modeller import parallel
job = parallel.job(parallel.local_slave()
for _ in range(ncpu))
a.use_parallel_job(job)
a.make()
for output in a.outputs:
cmd.load(output['name'], name, quiet=quiet)
finally:
os.chdir(cwd)
shutil.rmtree(tempdir)
cmd.align(name, selection, cycles=0)
if not sequence:
update_identifiers(name, selection)
if not quiet:
print(' peptide_rebuild_modeller: done')
def calculate_modeller_score(self, res_path):
"""
* This function constructs a single comparative model for the query
sequence from the known template structure, using alignment.ali,
a PIR format alignment of query and template. The final model is
written into the PDB file.
* This function also returns the DOPE assessed score of the model
generated by MODELLER. It actually returns the opposite (multiplied by -1) since it
is an energy score. This is to simplify the min/max normalization done afterwards.
DOPE is the most reliable score at separating native-like models from decoys (lower,
i.e, more negative, DOPE scores tend to correlate with more native-like models).
Args:
res_path (str): Path to the results folder.
Returns:
score(float): The DOPE score (multiplied by -1) of the model generated by MODELLER.
"""
root_dir = os.getcwd()
modeller_out_dir = res_path + "/modeller/"
ali_dir = "alignments/"
pathlib.Path(modeller_out_dir + ali_dir).mkdir(parents=True,
exist_ok=True)
# MODELLER generates the result files in his current directory, so we must
# go to the results directory and come back to root dir afterwards.
os.chdir(modeller_out_dir)
path_to_atm = root_dir + "/data/pdb/" + self.template.name
# We reindex all the PDB files to avoid any problem with modeller
self.template.reindex_pdb(1, path_to_atm, True)
# Parse the new PDB to get new residues and their coordinates generated by MODELLER
self.template.parse_pdb(path_to_atm + "/" +
self.template.reindexed_pdb + ".atm")
# Write Modeller's alignment PIR file
self.write_alignment_for_modeller("./alignments/")
# Redirect Modeller's verbose into nothingness, nil, chaos and abysses !
with contextlib.redirect_stdout(None):
# create a new MODELLER environment to build this model in
m.env = m.environ()
# directories for input atom files
m.env.io.atom_files_directory = [path_to_atm]
a_model = am.automodel(
m.env,
# alignment filename
alnfile=ali_dir + self.template.name + '.ali',
# codes of the templates
knowns=self.template.reindexed_pdb,
# code of the target
sequence='query_' + self.template.name,
# DOPEHR is very similar to DOPEHR but is obtained at
# Higher Resolution (using a bin size of 0.125Å
# rather than 0.5Å).
assess_methods=assess.DOPEHR)
a_model.very_fast()
# index of the first and last model (determines how many models to calculate)
a_model.starting_model = 1
a_model.ending_model = 1
modeller_dope_score = 0
# Catch any errors that Modeller can raise and write them in the log file
try:
a_model.make()
except m.ModellerError as err:
logging.warning(
"Modeller error with " + self.template.name + " | " +
self.template.pdb, str(err))
new_model_pdb = a_model.outputs[0]["name"]
modeller_dope_score = a_model.outputs[0]["DOPE-HR score"]
self.template.modeller_pdb = self.template.pdb + "_mod"
# Move the new model to the PDB directory and rename it
os.rename(new_model_pdb,
path_to_atm + "/" + self.template.modeller_pdb + ".atm")
# Parse the new model generated by MODELLER to get the residues and their coordinates
self.template.parse_pdb(path_to_atm + "/" +
self.template.modeller_pdb + ".atm")
# Go back to root directory
os.chdir(root_dir)
return modeller_dope_score * (-1)
def __run_modeller(self, alignFile, loopRefinement):
""".
Parameters
----------
alignFile : string
File containing the input data
result : list
The successfully calculated models are stored in this list
loopRefinement : boolean
If `True`, perform loop refinements
Returns
-------
list
Successfully calculated models
"""
log.none() # instructs Modeller to display no log output.
env = environ(
) # create a new MODELLER environment to build this model in
# Directories for input atom files
env.io.atom_files_directory = [
str(self.filePath.rstrip("/")),
]
env.schedule_scale = physical.values(default=1.0, soft_sphere=0.7)
# Selected atoms do not feel the neighborhood
# env.edat.nonbonded_sel_atoms = 2
env.io.hetatm = True # read in HETATM records from template PDBs
env.io.water = True # read in WATER records (including waters marked as HETATMs)
logger.debug(
"Performing loop refinement in addition to regular modelling: {}".
format(loopRefinement))
if not loopRefinement:
a = automodel(
env,
# alignment filename
alnfile=str(alignFile),
# codes of the templates
knowns=(str(self.templateID)),
# code of the target
sequence=str(self.seqID),
# wich method for validation should be calculated
assess_methods=(assess.DOPE, assess.normalized_dope),
)
else:
a = dope_loopmodel(
env,
# alignment filename
alnfile=str(alignFile),
# codes of the templates
knowns=(str(self.templateID)),
# code of the target
sequence=str(self.seqID),
# wich method for validation should be calculated
assess_methods=(assess.DOPE, assess.normalized_dope),
loop_assess_methods=(assess.DOPE, assess.normalized_dope),
)
# index of the first loop model
a.loop.starting_model = self.loopStart
# index of the last loop model
a.loop.ending_model = self.loopEnd
# loop refinement method; this yields
a.loop.md_level = refine.slow
a.starting_model = self.start # index of the first model
a.ending_model = self.end # index of the last model
# Very thorough VTFM optimization:
a.library_schedule = autosched.slow
a.max_var_iterations = 300
# Thorough MD optimization:
# a.md_level = refine.slow
a.md_level = None
# Repeat the whole cycle 2 times and do not stop unless obj.func. > 1E6
# a.repeat_optimization = 2
a.max_molpdf = 2e5
# with helper.print_heartbeats(): # use 'long_wait' in .travis.yml
with helper.log_print_statements(logger):
a.make() # do the actual homology modeling
# The output produced by modeller is stored in a.loop.outputs or a.outputs
# it is a dictionary
# Check for each model if it was successfully calculated, i.e.
# for each "normal" model and each loop model and append the
# assessment score to a list which is used to return the best model
result = []
loop = False
failures = []
# Add the normal output
for i in range(len(a.outputs)):
if not a.outputs[i]["failure"]:
model_filename = a.outputs[i]["name"]
model_dope_score = a.outputs[i]["Normalized DOPE score"]
logger.debug(
"Success! model_filename: {}, model_dope_score: {}".format(
model_filename, model_dope_score))
result.append((model_filename, model_dope_score))
else:
failure = a.outputs[i]["failure"]
logger.debug("Failure! {}".format(failure))
failures.append(a.outputs[i]["failure"])
# Add the loop refinement output
if loopRefinement:
logger.debug("Modeller loop outputs:")
for i in range(len(a.loop.outputs)):
if not a.loop.outputs[i]["failure"]:
model_filename = a.loop.outputs[i]["name"]
model_dope_score = a.loop.outputs[i][
"Normalized DOPE score"]
logger.debug(
"Success! model_filename: {}, model_dope_score: {}".
format(model_filename, model_dope_score))
result.append((model_filename, model_dope_score))
loop = True
else:
failure = a.loop.outputs[i]["failure"]
logger.debug("Failure! {}".format(failure))
failures.append(failure)
# Return the successfully calculated models and a loop flag indicating
# whether the returned models are loop refined or not
return result, loop, failures