def align_strs_seq(env, segfile, alnfile, knowns, sequence, matrix_file,
overhang=0, write_fit=False):
"""Align a single sequence with several structures"""
# Read the sequences of structures from the specs in SEGFILE:
aln = alignment(env, file=segfile, align_codes=knowns)
# Only align structures if there's more than one:
if len(aln) > 1:
# do a multiple sequence alignment of structures:
aln.malign(gap_penalties_1d=(-600, -400), overhang=overhang)
# do a multiple structural alignment of structures:
aln.malign3d(gap_penalties_3d=(0.0, 2.0), fit_atoms='CA',
overhang=overhang, write_fit=write_fit)
# remember the number of structures
align_block = len(aln)
# add the sequence of the unknown to the sequence/alignment arrays
# containing the aligned structures:
aln.append(file=segfile, align_codes=sequence)
# align the last sequence with the fixed alignment of structures:
aln.align(align_block=align_block, gap_penalties_1d=(-600, -400),
overhang=overhang)
# write the alignment to a file ALNFILE
aln.write(file=alnfile)
# do some sequence comparisons:
aln.id_table(matrix_file)
return aln
def sequence_srch(env,
sequence,
segfile,
chains_list,
toplib='${LIB}/top_heav.lib',
search_randomizations=0,
signif_cutoff=(4.0, 5.0)):
"""For a given target sequence, find all template chains in PDB.
- align all of them together and write the alignment to a file.
- calculate identity matrix and write it to a file.
- calculate a dendrogram and write it to the .log file."""
penalties = (-600, -400) # This should be fine for significance tests with
# default matrix as1.sim.mat
overhang = 20
seqfile = '$(LIB)/pdball.pir'
sdb = sequence_db(env,
seq_database_file=seqfile,
chains_list=chains_list,
seq_database_format='PIR')
aln = alignment(env, file=segfile, align_codes=sequence)
sdb.search(aln,
search_top_list=30,
off_diagonal=9999,
search_group_list='$(LIB)/pdb_40.grp',
seq_database_file=seqfile,
gap_penalties_1d=penalties,
output='SHORT',
signif_cutoff=signif_cutoff,
search_randomizations=search_randomizations)
if len(aln) > 1 or (len(aln) == 1 and aln[0].code != sequence):
aln.write(file='alignment.tmp')
aln.malign(overhang=overhang,
off_diagonal=150,
gap_penalties_1d=penalties)
aln.malign3d(overhang=overhang,
gap_penalties_3d=(0, 3),
off_diagonal=150)
blk = len(aln)
aln.append(file=segfile, align_codes=sequence)
# Only needed for ALIGN2D's PSA run
env.libs.topology.read(file=toplib)
aln.align2d(align_block=blk,
gap_penalties_1d=(-450, 0),
gap_penalties_2d=(0.35, 1.2, 0.9, 1.2, 0.6, 8.6, 1.2, 0,
0),
max_gap_length=50,
off_diagonal=150,
overhang=overhang)
aln.write(file=sequence + '.ali')
aln.write(file=sequence + '.pap', alignment_format='PAP',
alignment_features='HELIX BETA ACCESSIBILITY ' + \
'STRAIGHTNESS CONSERVATION INDICES')
mat = sequence + '.mat'
aln.id_table(matrix_file=mat)
env.dendrogram(matrix_file=mat, cluster_cut=-1.0)
def cluster(self, cluster_cut=1.5):
"""Cluster all output models, and output an optimized cluster average"""
self.read_initial_model()
aln = alignment(self.env)
self.align_models(aln)
if len(aln) == 0:
log.error('cluster', 'No generated models - nothing to cluster!')
aln.malign3d(gap_penalties_3d=(0, 3), fit=False)
aln.append_model(mdl=self,
align_codes='cluster',
atom_files='cluster.opt')
self.transfer_xyz(aln, cluster_cut=cluster_cut)
self.write(file='cluster.ini')
self.read_top_par()
self.rd_restraints()
self.create_topology(aln, sequence='cluster')
atmsel = self._check_select_atoms()
self.restraints.unpick_all()
self.restraints.pick(atmsel)
self.restraints.condense()
edat = energy_data(copy=self.env.edat)
edat.nonbonded_sel_atoms = 1
atmsel.energy(output='LONG', edat=edat)
cg = conjugate_gradients()
cg.optimize(atmsel,
actions=actions.trace(5, 'cluster.deb'),
max_iterations=self.max_var_iterations)
atmsel.energy()
self.write(file='cluster.opt')
aln.compare_structures(fit=True)
def refine(self, atmsel, actions):
"""Refine the optimized model with MD and CG"""
# Save the current model:
if self.fit_in_refine != 'NO_FIT':
self.write(file='TO_BE_REFINED.TMP')
# Possibly skip selecting hot atoms only and optimize all atoms:
if self.refine_hot_only:
self.initial_refine_hot(atmsel)
# Do simulated annealing MD:
if self.md_level:
self.md_level(atmsel, actions)
# Possibly skip 'HOT CG' after MD:
if self.refine_hot_only:
self.final_refine_hot(atmsel)
# Get a final conjugate gradients refined structure:
cg = conjugate_gradients()
cg.optimize(atmsel,
max_iterations=200,
output=self.optimize_output,
actions=actions)
# Evaluate gross changes between the initial and final refined model:
if 'NO_FIT' not in self.fit_in_refine:
aln = alignment(self.env)
mdl2 = read_model(file='TO_BE_REFINED.TMP')
casel = selection(self).only_atom_types('CA')
casel.superpose(mdl2, aln)
casel = selection(self)
casel.superpose(mdl2, aln)
modfile.delete('TO_BE_REFINED.TMP')
def read_alignment(self, aln=None):
"""Read the template-sequence alignment needed for modeling"""
if aln is None:
aln = alignment(self.env)
aln.clear()
aln.append(file=self.alnfile,
align_codes=self.knowns + [self.sequence])
return aln
def make(self,
atmsel,
restraint_type,
spline_on_site,
residue_span_range=(0, 99999),
residue_span_sign=True,
restraint_sel_atoms=1,
basis_pdf_weight='LOCAL',
basis_relative_weight=0.05,
intersegment=True,
dih_lib_only=False,
spline_dx=0.5,
spline_min_points=5,
spline_range=4.0,
mnch_lib=1,
accessibility_type=8,
surftyp=1,
distngh=6.0,
aln=None,
edat=None,
io=None):
"""Calculates and selects new restraints of a specified type"""
if not hasattr(atmsel, "get_atom_indices"):
raise TypeError("First argument needs to be an atom selection")
if not isinstance(restraint_type, str):
raise TypeError("restraint_type must be a string")
restyp = restraint_type.upper()
(inds, mdl) = atmsel.get_atom_indices()
if mdl is None:
raise ValueError("selection is empty")
if mdl is not self.__mdl:
raise ValueError("selection refers to a different model")
if edat is None:
edat = self.__mdl.env.edat
if io is None:
io = self.__mdl.env.io
if aln is None:
if restyp in \
('CHI1_DIHEDRAL', 'CHI2_DIHEDRAL', 'CHI3_DIHEDRAL',
'CHI4_DIHEDRAL', 'PHI_DIHEDRAL', 'PSI_DIHEDRAL',
'OMEGA_DIHEDRAL', 'PHI-PSI_BINORMAL'):
raise ValueError("You must provide an alignment for this " +
"restraint type")
else:
aln = alignment.alignment(self.__mdl.env)
func = _modeller.mod_restraints_make
return func(self.__mdl.modpt, edat.modpt, aln.modpt, io.modpt,
self.__mdl.env.libs.modpt, inds, (), (),
residue_span_range, restraint_type, restraint_sel_atoms, 0,
basis_pdf_weight, 0, 0., basis_relative_weight,
spline_on_site, residue_span_sign, intersegment,
dih_lib_only, spline_dx, spline_min_points, spline_range,
mnch_lib, accessibility_type, (0, 0), (0, 0, 0), surftyp,
distngh, 0.0)
def principal_components(env, family, cluster_cut):
"""Does principal components on family"""
aln = alignment(env, file=family + '.ali')
mat = family + '.mat'
aln.id_table(matrix_file=mat)
env.principal_components(matrix_file=mat, file=family + '.dat')
env.dendrogram(matrix_file=mat, cluster_cut=cluster_cut)
def fit(env, model, code, model2, code2, alnfile, model2_fit):
"""Superposes model2 on model, and writes out a file with model2 superposed
on model."""
m1 = modelobj(env, file=model)
m2 = modelobj(env, file=model2)
aln = alignment(env, file=alnfile, align_codes=(code, code2))
atmsel = selection(m1).only_atom_types('CA')
atmsel.superpose(m2, aln)
m2.write(file=model2_fit)
def dihedrals(atmsel):
"""Randomize dihedrals"""
mdl = atmsel.get_model()
aln = alignment(mdl.env, file=mdl.alnfile,
align_codes=mdl.knowns+[mdl.sequence])
# Just in case, generate topology again (ROTATE needs bonds)
# (could replace with GENERATE_TOPOLOGY if no special patches)
mdl.create_topology(aln)
# Optimize all dihedral angles:
atmsel.rotate_dihedrals(change='RANDOMIZE', deviation=mdl.deviation,
dihedrals=('phi', 'psi', 'omega', 'chi1', 'chi2',
'chi3', 'chi4'))
def build_sequence(self,
sequence,
special_patches=None,
patch_default=True,
blank_single_chain=True):
"""Build an extended chain from a string of one-letter residue codes"""
a = alignment.alignment(self.env)
a.append_sequence(sequence, blank_single_chain)
self.clear_topology()
self.generate_topology(a[0],
patch_default=patch_default,
blank_single_chain=blank_single_chain)
if special_patches:
special_patches(self)
self.build(initialize_xyz=True, build_method='INTERNAL_COORDINATES')
self.prottyp = 'structure'
def read_xyz(mdl, aln):
"""Read in the initial structure from an existing file"""
# Create two copies of the template sequence:
a = alignment(mdl.env)
a.append(file=mdl.alnfile, align_codes=[mdl.sequence]*2)
# Use the initial model as the first structure in the alignment:
a[0].prottyp = 'structureX'
a[0].atom_file = a[0].code = mdl.my_inifile
mdl.read_top_par()
mdl.create_topology(aln)
# Get model coordinates from the initial model, making sure that the
# sequence is correct and any missing atoms are filled in:
mdl.transfer_xyz(a, cluster_cut=-1.0)
mdl.build(initialize_xyz=False, build_method='INTERNAL_COORDINATES')
def fit_models_on_template(self):
"""Superpose each of the generated models on the templates"""
aln = alignment(self.env)
aln.append(file=self.alnfile, align_codes=self.knowns)
self.align_models(aln)
# To take care of the '.' in segment specs:
aln.write(file='.tmp.ali', alignment_format='PIR')
codes = [seq.code for seq in aln]
aln.read(file='.tmp.ali', alignment_format='PIR', align_codes=codes)
modfile.delete('.tmp.ali')
aln.compare_structures(fit=True, output='SHORT', fit_atoms='CA')
aln.malign3d(gap_penalties_3d=(0, 3),
write_whole_pdb=False,
write_fit=True,
fit=False,
fit_atoms='CA',
current_directory=True,
edit_file_ext=(self.pdb_ext, '_fit.pdb'),
output='SHORT')
