def simulate(file="simulation.pdb.bz2", dir=None, step_size=2.0, snapshot=10, total=1000, model=1, force=True):
"""Pseudo-Brownian dynamics simulation of the frame order motions.
@keyword file: The PDB file for storing the frame order pseudo-Brownian dynamics simulation. The compression is determined automatically by the file extensions '*.pdb', '*.pdb.gz', and '*.pdb.bz2'.
@type file: str
@keyword dir: The directory name to place the file into.
@type dir: str or None
@keyword step_size: The rotation will be of a random direction but with this fixed angle. The value is in degrees.
@type step_size: float
@keyword snapshot: The number of steps in the simulation when snapshots will be taken.
@type snapshot: int
@keyword total: The total number of snapshots to take before stopping the simulation.
@type total: int
@keyword model: Only one model from an analysed ensemble of structures can be used for the pseudo-Brownian simulation, as the simulation and corresponding PDB file consists of one model per simulation.
@type model: int
@keyword force: A flag which, if set to True, will overwrite the any pre-existing file.
@type force: bool
"""
# Printout.
print("Pseudo-Brownian dynamics simulation of the frame order motions.")
# Checks.
check_pipe()
check_model()
check_domain()
check_parameters()
check_pivot()
# Skip the rigid model.
if cdp.model == MODEL_RIGID:
print("Skipping the rigid model.")
return
# Open the output file.
file = open_write_file(file_name=file, dir=dir, force=force)
# The parameter values.
values = assemble_param_vector()
params = {}
i = 0
for name in cdp.params:
params[name] = values[i]
i += 1
# The structure.
structure = deepcopy(cdp.structure)
if structure.num_models() > 1:
structure.collapse_ensemble(model_num=model)
# The pivot points.
num_states = 1
if cdp.model == MODEL_DOUBLE_ROTOR:
num_states = 2
pivot = zeros((num_states, 3), float64)
for i in range(num_states):
pivot[i] = generate_pivot(order=i+1, pdb_limit=True)
# Shift to the average position.
average_position(structure=structure, models=[None])
# The motional eigenframe.
frame = generate_axis_system()
# Create the distribution.
brownian(file=file, model=cdp.model, structure=structure, parameters=params, eigenframe=frame, pivot=pivot, atom_id=domain_moving(), step_size=step_size, snapshot=snapshot, total=total)
# Close the file.
file.close()
def distribute(file="distribution.pdb.bz2", dir=None, atom_id=None, total=1000, max_rotations=100000, model=1, force=True):
"""Create a uniform distribution of structures for the frame order motions.
@keyword file: The PDB file for storing the frame order motional distribution. The compression is determined automatically by the file extensions '*.pdb', '*.pdb.gz', and '*.pdb.bz2'.
@type file: str
@keyword dir: The directory name to place the file into.
@type dir: str or None
@keyword atom_id: The atom identification string to allow the distribution to be a subset of all atoms.
@type atom_id: None or str
@keyword total: The total number of states/model/structures in the distribution.
@type total: int
@keyword max_rotations: The maximum number of rotations to generate the distribution from. This prevents an execution for an infinite amount of time when a frame order amplitude parameter is close to zero so that the subset of all rotations within the distribution is close to zero.
@type max_rotations: int
@keyword model: Only one model from an analysed ensemble of structures can be used for the distribution, as the corresponding PDB file consists of one model per state.
@type model: int
@keyword force: A flag which, if set to True, will overwrite the any pre-existing file.
@type force: bool
"""
# Printout.
print("Uniform distribution of structures representing the frame order motions.")
# Check the total.
if total > 9999:
raise RelaxError("A maximum of 9999 models is allowed in the PDB format.")
# Checks.
check_pipe()
check_model()
check_domain()
check_parameters()
check_pivot()
# Skip the rigid model.
if cdp.model == MODEL_RIGID:
print("Skipping the rigid model.")
return
# Open the output file.
file = open_write_file(file_name=file, dir=dir, force=force)
# The parameter values.
values = assemble_param_vector()
params = {}
i = 0
for name in cdp.params:
params[name] = values[i]
i += 1
# The structure.
structure = deepcopy(cdp.structure)
if structure.num_models() > 1:
structure.collapse_ensemble(model_num=model)
# The pivot points.
num_states = 1
if cdp.model == MODEL_DOUBLE_ROTOR:
num_states = 2
pivot = zeros((num_states, 3), float64)
for i in range(num_states):
pivot[i] = generate_pivot(order=i+1, pdb_limit=True)
# Shift to the average position.
average_position(structure=structure, models=[None])
# The motional eigenframe.
frame = generate_axis_system()
# Only work with a subset.
if atom_id:
# The inverted selection.
selection = structure.selection(atom_id=atom_id, inv=True)
# Delete the data.
structure.delete(selection=selection, verbosity=0)
# Create the distribution.
uniform_distribution(file=file, model=cdp.model, structure=structure, parameters=params, eigenframe=frame, pivot=pivot, atom_id=domain_moving(), total=total, max_rotations=max_rotations)
# Close the file.
file.close()
def distribute(file="distribution.pdb.bz2",
dir=None,
atom_id=None,
total=1000,
max_rotations=100000,
model=1,
force=True):
"""Create a uniform distribution of structures for the frame order motions.
@keyword file: The PDB file for storing the frame order motional distribution. The compression is determined automatically by the file extensions '*.pdb', '*.pdb.gz', and '*.pdb.bz2'.
@type file: str
@keyword dir: The directory name to place the file into.
@type dir: str or None
@keyword atom_id: The atom identification string to allow the distribution to be a subset of all atoms.
@type atom_id: None or str
@keyword total: The total number of states/model/structures in the distribution.
@type total: int
@keyword max_rotations: The maximum number of rotations to generate the distribution from. This prevents an execution for an infinite amount of time when a frame order amplitude parameter is close to zero so that the subset of all rotations within the distribution is close to zero.
@type max_rotations: int
@keyword model: Only one model from an analysed ensemble of structures can be used for the distribution, as the corresponding PDB file consists of one model per state.
@type model: int
@keyword force: A flag which, if set to True, will overwrite the any pre-existing file.
@type force: bool
"""
# Printout.
print(
"Uniform distribution of structures representing the frame order motions."
)
# Check the total.
if total > 9999:
raise RelaxError(
"A maximum of 9999 models is allowed in the PDB format.")
# Checks.
check_pipe()
check_model()
check_domain()
check_parameters()
check_pivot()
# Skip the rigid model.
if cdp.model == MODEL_RIGID:
print("Skipping the rigid model.")
return
# Open the output file.
file = open_write_file(file_name=file, dir=dir, force=force)
# The parameter values.
values = assemble_param_vector()
params = {}
i = 0
for name in cdp.params:
params[name] = values[i]
i += 1
# The structure.
structure = deepcopy(cdp.structure)
if structure.num_models() > 1:
structure.collapse_ensemble(model_num=model)
# The pivot points.
num_states = 1
if cdp.model == MODEL_DOUBLE_ROTOR:
num_states = 2
pivot = zeros((num_states, 3), float64)
for i in range(num_states):
pivot[i] = generate_pivot(order=i + 1, pdb_limit=True)
# Shift to the average position.
average_position(structure=structure, models=[None])
# The motional eigenframe.
frame = generate_axis_system()
# Only work with a subset.
if atom_id:
# The inverted selection.
selection = structure.selection(atom_id=atom_id, inv=True)
# Delete the data.
structure.delete(selection=selection, verbosity=0)
# Create the distribution.
uniform_distribution(file=file,
model=cdp.model,
structure=structure,
parameters=params,
eigenframe=frame,
pivot=pivot,
atom_id=domain_moving(),
total=total,
max_rotations=max_rotations)
# Close the file.
file.close()
def simulate(file="simulation.pdb.bz2",
dir=None,
step_size=2.0,
snapshot=10,
total=1000,
model=1,
force=True):
"""Pseudo-Brownian dynamics simulation of the frame order motions.
@keyword file: The PDB file for storing the frame order pseudo-Brownian dynamics simulation. The compression is determined automatically by the file extensions '*.pdb', '*.pdb.gz', and '*.pdb.bz2'.
@type file: str
@keyword dir: The directory name to place the file into.
@type dir: str or None
@keyword step_size: The rotation will be of a random direction but with this fixed angle. The value is in degrees.
@type step_size: float
@keyword snapshot: The number of steps in the simulation when snapshots will be taken.
@type snapshot: int
@keyword total: The total number of snapshots to take before stopping the simulation.
@type total: int
@keyword model: Only one model from an analysed ensemble of structures can be used for the pseudo-Brownian simulation, as the simulation and corresponding PDB file consists of one model per simulation.
@type model: int
@keyword force: A flag which, if set to True, will overwrite the any pre-existing file.
@type force: bool
"""
# Printout.
print("Pseudo-Brownian dynamics simulation of the frame order motions.")
# Checks.
check_pipe()
check_model()
check_domain()
check_parameters()
check_pivot()
# Skip the rigid model.
if cdp.model == MODEL_RIGID:
print("Skipping the rigid model.")
return
# Open the output file.
file = open_write_file(file_name=file, dir=dir, force=force)
# The parameter values.
values = assemble_param_vector()
params = {}
i = 0
for name in cdp.params:
params[name] = values[i]
i += 1
# The structure.
structure = deepcopy(cdp.structure)
if structure.num_models() > 1:
structure.collapse_ensemble(model_num=model)
# The pivot points.
num_states = 1
if cdp.model == MODEL_DOUBLE_ROTOR:
num_states = 2
pivot = zeros((num_states, 3), float64)
for i in range(num_states):
pivot[i] = generate_pivot(order=i + 1, pdb_limit=True)
# Shift to the average position.
average_position(structure=structure, models=[None])
# The motional eigenframe.
frame = generate_axis_system()
# Create the distribution.
brownian(file=file,
model=cdp.model,
structure=structure,
parameters=params,
eigenframe=frame,
pivot=pivot,
atom_id=domain_moving(),
step_size=step_size,
snapshot=snapshot,
total=total)
# Close the file.
file.close()
def decompose(root="decomposed", dir=None, atom_id=None, model=1, force=True):
"""Structural representation of the individual frame order motional components.
@keyword root: The file root for the PDB files created. Each motional component will be represented by a different PDB file appended with '_mode1.pdb', '_mode2.pdb', '_mode3.pdb', etc.
@type root: str
@keyword dir: The directory name to place the file into.
@type dir: str or None
@keyword atom_id: The atom identification string to allow the decomposition to be applied to subset of all atoms.
@type atom_id: None or str
@keyword model: Only one model from an analysed ensemble of structures can be used for the decomposition, as the corresponding PDB file consists of one model per state.
@type model: int
@keyword force: A flag which, if set to True, will overwrite the any pre-existing file.
@type force: bool
"""
# Printout.
print(
"PDB representation of the individual components of the frame order motions."
)
# Checks.
check_pipe()
check_model()
check_domain()
check_parameters()
check_pivot()
# Skip any unsupported models.
unsupported = [MODEL_RIGID, MODEL_DOUBLE_ROTOR]
if cdp.model in unsupported:
print("Skipping the unsupported '%s' model." % cdp.model)
return
# Initialise the angle vector (cone opening angle 1, cone opening angle 2, torsion angle).
angles = zeros(3, float64)
# Cone opening.
if cdp.model in MODEL_LIST_ISO_CONE:
angles[0] = angles[1] = cdp.cone_theta
elif cdp.model in MODEL_LIST_PSEUDO_ELLIPSE:
angles[0] = cdp.cone_theta_y
angles[1] = cdp.cone_theta_x
# Non-zero torsion angle.
if cdp.model in MODEL_LIST_FREE_ROTORS:
angles[2] = pi
elif cdp.model in MODEL_LIST_RESTRICTED_TORSION:
angles[2] = cdp.cone_sigma_max
# The motional eigenframe.
frame = generate_axis_system()
# Mode ordering from largest to smallest.
indices = argsort(angles)
angles = angles[indices[::-1]]
frame = transpose(transpose(frame)[indices[::-1]])
# The pivot point.
pivot = generate_pivot(order=1, pdb_limit=True)
# Loop over each mode.
for i in range(3):
# Skip modes with no motion.
if angles[i] < 1e-7:
continue
# Open the output file.
file_name = "%s_mode%i.pdb" % (root, i + 1)
file = open_write_file(file_name=file_name, dir=dir, force=force)
# The structure.
structure = deepcopy(cdp.structure)
if structure.num_models() > 1:
structure.collapse_ensemble(model_num=model)
# Shift to the average position.
average_position(structure=structure, models=[None])
# Create the representation.
mode_distribution(file=file,
structure=structure,
axis=frame[:, i],
angle=angles[i],
pivot=pivot,
atom_id=domain_moving())
# Close the file.
file.close()
def generate_pivot(order=1, sim_index=None, pipe_name=None, pdb_limit=False):
"""Create and return the given pivot.
@keyword order: The pivot number with 1 corresponding to the first pivot, 2 to the second, etc.
@type order: int
@keyword sim_index: The optional Monte Carlo simulation index. If provided, the pivot for the given simulation will be returned instead.
@type sim_index: None or int
@keyword pipe_name: The data pipe
@type pipe_name: str
@keyword pdb_limit: A flag which if True will cause the coordinate to be between -1000 and 1000.
@type pdb_limit: bool
@return: The give pivot point.
@rtype: numpy 3D rank-1 float64 array
"""
# Checks.
check_pipe(pipe_name)
check_pivot(pipe_name=pipe_name)
check_model(pipe_name=pipe_name)
# The data pipe.
if pipe_name == None:
pipe_name = pipes.cdp_name()
# Get the data pipe.
dp = pipes.get_pipe(pipe_name)
# Initialise.
pivot = None
# The double rotor parameterisation.
if dp.model in [MODEL_DOUBLE_ROTOR]:
# The 2nd pivot point (the centre of the frame).
if sim_index != None and hasattr(dp, 'pivot_x_sim'):
pivot_2nd = array([dp.pivot_x_sim[sim_index], dp.pivot_y_sim[sim_index], dp.pivot_z_sim[sim_index]], float64)
else:
pivot_2nd = array([dp.pivot_x, dp.pivot_y, dp.pivot_z], float64)
# Generate the first pivot.
if order == 1:
# The eigenframe.
frame = zeros((3, 3), float64)
if sim_index != None and hasattr(dp, 'pivot_disp_sim'):
euler_to_R_zyz(dp.eigen_alpha_sim[sim_index], dp.eigen_beta_sim[sim_index], dp.eigen_gamma_sim[sim_index], frame)
pivot_disp = dp.pivot_disp_sim[sim_index]
else:
euler_to_R_zyz(dp.eigen_alpha, dp.eigen_beta, dp.eigen_gamma, frame)
pivot_disp = dp.pivot_disp
# The 1st pivot.
pivot = pivot_2nd + frame[:, 2] * pivot_disp
# Alias the 2nd pivot.
elif order == 2:
pivot = pivot_2nd
# All other models.
elif order == 1:
if sim_index != None and hasattr(dp, 'pivot_x_sim'):
pivot = array([dp.pivot_x_sim[sim_index], dp.pivot_y_sim[sim_index], dp.pivot_z_sim[sim_index]], float64)
else:
pivot = array([dp.pivot_x, dp.pivot_y, dp.pivot_z], float64)
# PDB limits.
if pivot is not None and pdb_limit:
# The original pivot, as text.
orig_pivot = "[%.3f, %.3f, %.3f]" % (pivot[0], pivot[1], pivot[2])
# Check each coordinate.
out = False
for i in range(3):
if pivot[i] <= -900.0:
pivot[i] = -900.0
out = True
elif pivot[i] > 9900.0:
pivot[i] = 9900.0
out = True
# Failure.
if out:
new_pivot = "[%.3f, %.3f, %.3f]" % (pivot[0], pivot[1], pivot[2])
warn(RelaxWarning("The pivot point %s is outside of the PDB coordinate limits of [-999.999, 9999.999], less a 100 Angstrom buffer, shifting to %s." % (orig_pivot, new_pivot)))
# Return the pivot.
return pivot
def grid_search(self,
lower=None,
upper=None,
inc=None,
scaling_matrix=None,
constraints=False,
verbosity=0,
sim_index=None):
"""Perform a grid search.
@keyword lower: The per-model lower bounds of the grid search which must be equal to the number of parameters in the model.
@type lower: list of lists of numbers
@keyword upper: The per-model upper bounds of the grid search which must be equal to the number of parameters in the model.
@type upper: list of lists of numbers
@keyword inc: The per-model increments for each dimension of the space for the grid search. The number of elements in the array must equal to the number of parameters in the model.
@type inc: list of lists of int
@keyword scaling_matrix: The per-model list of diagonal and square scaling matrices.
@type scaling_matrix: list of numpy rank-2, float64 array or list of None
@keyword constraints: If True, constraints are applied during the grid search (eliminating parts of the grid). If False, no constraints are used.
@type constraints: bool
@keyword verbosity: A flag specifying the amount of information to print. The higher the value, the greater the verbosity.
@type verbosity: int
@keyword sim_index: The Monte Carlo simulation index.
@type sim_index: None or int
"""
# Test if the Frame Order model has been set up.
if not hasattr(cdp, 'model'):
raise RelaxNoModelError('Frame Order')
# Test if the pivot has been set.
check_pivot()
# The number of parameters.
n = param_num()
# Alias the single model grid bounds and increments.
lower = lower[0]
upper = upper[0]
inc = inc[0]
# Initialise the grid increments structures.
grid = []
"""This structure is a list of lists. The first dimension corresponds to the model
parameter. The second dimension are the grid node positions."""
# Generate the grid.
for i in range(n):
# Fixed parameter.
if inc[i] == None:
grid.append(None)
continue
# Reset.
dist_type = None
end_point = True
# Arccos grid from 0 to pi.
if cdp.params[i] in ['ave_pos_beta', 'eigen_beta', 'axis_theta']:
# Change the default increment numbers.
if not isinstance(inc, list):
inc[i] = int(inc[i] / 2) + 1
# The distribution type and end point.
dist_type = 'acos'
end_point = False
# Append the grid row.
row = grid_row(inc[i],
lower[i],
upper[i],
dist_type=dist_type,
end_point=end_point)
grid.append(row)
# Remove an inc if the end point has been removed.
if not end_point:
inc[i] -= 1
# Total number of points.
total_pts = 1
for i in range(n):
# Fixed parameter.
if grid[i] == None:
continue
total_pts = total_pts * len(grid[i])
# Check the number.
max_pts = 50e6
if total_pts > max_pts:
raise RelaxError(
"The total number of grid points '%s' exceeds the maximum of '%s'."
% (total_pts, int(max_pts)))
# Build the points array.
pts = zeros((total_pts, n), float64)
indices = zeros(n, int)
for i in range(total_pts):
# Loop over the dimensions.
for j in range(n):
# Fixed parameter.
if grid[j] == None:
# Get the current parameter value.
pts[i, j] = getattr(
cdp, cdp.params[j]) / scaling_matrix[0][j, j]
# Add the point coordinate.
else:
pts[i, j] = grid[j][indices[j]] / scaling_matrix[0][j, j]
# Increment the step positions.
for j in range(n):
if inc[j] != None and indices[j] < inc[j] - 1:
indices[j] += 1
break # Exit so that the other step numbers are not incremented.
else:
indices[j] = 0
# Linear constraints.
A, b = None, None
if constraints:
# Obtain the constraints.
A, b = linear_constraints(scaling_matrix=scaling_matrix[0])
# Constraint flag set but no constraints present.
if A is None:
if verbosity:
warn(
RelaxWarning(
"The '%s' model parameters are not constrained, turning the linear constraint algorithm off."
% cdp.model))
constraints = False
# The numeric integration information.
if not hasattr(cdp, 'quad_int'):
cdp.quad_int = False
sobol_max_points, sobol_oversample = None, None
if hasattr(cdp, 'sobol_max_points'):
sobol_max_points = cdp.sobol_max_points
sobol_oversample = cdp.sobol_oversample
# Set up the data structures for the target function.
param_vector, full_tensors, full_in_ref_frame, rdcs, rdc_err, rdc_weight, rdc_vect, rdc_const, pcs, pcs_err, pcs_weight, atomic_pos, temp, frq, paramag_centre, com, ave_pos_pivot, pivot, pivot_opt = target_fn_data_setup(
sim_index=sim_index, verbosity=verbosity)
# Get the Processor box singleton (it contains the Processor instance) and alias the Processor.
processor_box = Processor_box()
processor = processor_box.processor
# Set up for multi-processor execution.
if processor.processor_size() > 1:
# Printout.
print("Parallelised grid search.")
print(
"Randomising the grid points to equalise the time required for each grid subdivision.\n"
)
# Randomise the points.
shuffle(pts)
# Loop over each grid subdivision, with all points violating constraints being eliminated.
for subdivision in grid_split_array(
divisions=processor.processor_size(),
points=pts,
A=A,
b=b,
verbosity=verbosity):
# Set up the memo for storage on the master.
memo = Frame_order_memo(sim_index=sim_index,
scaling_matrix=scaling_matrix[0])
# Set up the command object to send to the slave and execute.
command = Frame_order_grid_command(
points=subdivision,
scaling_matrix=scaling_matrix[0],
sim_index=sim_index,
model=cdp.model,
param_vector=param_vector,
full_tensors=full_tensors,
full_in_ref_frame=full_in_ref_frame,
rdcs=rdcs,
rdc_err=rdc_err,
rdc_weight=rdc_weight,
rdc_vect=rdc_vect,
rdc_const=rdc_const,
pcs=pcs,
pcs_err=pcs_err,
pcs_weight=pcs_weight,
atomic_pos=atomic_pos,
temp=temp,
frq=frq,
paramag_centre=paramag_centre,
com=com,
ave_pos_pivot=ave_pos_pivot,
pivot=pivot,
pivot_opt=pivot_opt,
sobol_max_points=sobol_max_points,
sobol_oversample=sobol_oversample,
verbosity=verbosity,
quad_int=cdp.quad_int)
# Add the slave command and memo to the processor queue.
processor.add_to_queue(command, memo)
# Execute the queued elements.
processor.run_queue()
def generate_pivot(order=1, sim_index=None, pipe_name=None, pdb_limit=False):
"""Create and return the given pivot.
@keyword order: The pivot number with 1 corresponding to the first pivot, 2 to the second, etc.
@type order: int
@keyword sim_index: The optional Monte Carlo simulation index. If provided, the pivot for the given simulation will be returned instead.
@type sim_index: None or int
@keyword pipe_name: The data pipe
@type pipe_name: str
@keyword pdb_limit: A flag which if True will cause the coordinate to be between -1000 and 1000.
@type pdb_limit: bool
@return: The give pivot point.
@rtype: numpy 3D rank-1 float64 array
"""
# Checks.
check_pipe(pipe_name)
check_pivot(pipe_name=pipe_name)
check_model(pipe_name=pipe_name)
# The data pipe.
if pipe_name == None:
pipe_name = pipes.cdp_name()
# Get the data pipe.
dp = pipes.get_pipe(pipe_name)
# Initialise.
pivot = None
# The double rotor parameterisation.
if dp.model in [MODEL_DOUBLE_ROTOR]:
# The 2nd pivot point (the centre of the frame).
if sim_index != None and hasattr(dp, 'pivot_x_sim'):
pivot_2nd = array([
dp.pivot_x_sim[sim_index], dp.pivot_y_sim[sim_index],
dp.pivot_z_sim[sim_index]
], float64)
else:
pivot_2nd = array([dp.pivot_x, dp.pivot_y, dp.pivot_z], float64)
# Generate the first pivot.
if order == 1:
# The eigenframe.
frame = zeros((3, 3), float64)
if sim_index != None and hasattr(dp, 'pivot_disp_sim'):
euler_to_R_zyz(dp.eigen_alpha_sim[sim_index],
dp.eigen_beta_sim[sim_index],
dp.eigen_gamma_sim[sim_index], frame)
pivot_disp = dp.pivot_disp_sim[sim_index]
else:
euler_to_R_zyz(dp.eigen_alpha, dp.eigen_beta, dp.eigen_gamma,
frame)
pivot_disp = dp.pivot_disp
# The 1st pivot.
pivot = pivot_2nd + frame[:, 2] * pivot_disp
# Alias the 2nd pivot.
elif order == 2:
pivot = pivot_2nd
# All other models.
elif order == 1:
if sim_index != None and hasattr(dp, 'pivot_x_sim'):
pivot = array([
dp.pivot_x_sim[sim_index], dp.pivot_y_sim[sim_index],
dp.pivot_z_sim[sim_index]
], float64)
else:
pivot = array([dp.pivot_x, dp.pivot_y, dp.pivot_z], float64)
# PDB limits.
if pivot is not None and pdb_limit:
# The original pivot, as text.
orig_pivot = "[%.3f, %.3f, %.3f]" % (pivot[0], pivot[1], pivot[2])
# Check each coordinate.
out = False
for i in range(3):
if pivot[i] <= -900.0:
pivot[i] = -900.0
out = True
elif pivot[i] > 9900.0:
pivot[i] = 9900.0
out = True
# Failure.
if out:
new_pivot = "[%.3f, %.3f, %.3f]" % (pivot[0], pivot[1], pivot[2])
warn(
RelaxWarning(
"The pivot point %s is outside of the PDB coordinate limits of [-999.999, 9999.999], less a 100 Angstrom buffer, shifting to %s."
% (orig_pivot, new_pivot)))
# Return the pivot.
return pivot
def grid_search(self, lower=None, upper=None, inc=None, scaling_matrix=None, constraints=False, verbosity=0, sim_index=None):
"""Perform a grid search.
@keyword lower: The per-model lower bounds of the grid search which must be equal to the number of parameters in the model.
@type lower: list of lists of numbers
@keyword upper: The per-model upper bounds of the grid search which must be equal to the number of parameters in the model.
@type upper: list of lists of numbers
@keyword inc: The per-model increments for each dimension of the space for the grid search. The number of elements in the array must equal to the number of parameters in the model.
@type inc: list of lists of int
@keyword scaling_matrix: The per-model list of diagonal and square scaling matrices.
@type scaling_matrix: list of numpy rank-2, float64 array or list of None
@keyword constraints: If True, constraints are applied during the grid search (eliminating parts of the grid). If False, no constraints are used.
@type constraints: bool
@keyword verbosity: A flag specifying the amount of information to print. The higher the value, the greater the verbosity.
@type verbosity: int
@keyword sim_index: The Monte Carlo simulation index.
@type sim_index: None or int
"""
# Test if the Frame Order model has been set up.
if not hasattr(cdp, 'model'):
raise RelaxNoModelError('Frame Order')
# Test if the pivot has been set.
check_pivot()
# The number of parameters.
n = param_num()
# Alias the single model grid bounds and increments.
lower = lower[0]
upper = upper[0]
inc = inc[0]
# Initialise the grid increments structures.
grid = []
"""This structure is a list of lists. The first dimension corresponds to the model
parameter. The second dimension are the grid node positions."""
# Generate the grid.
for i in range(n):
# Fixed parameter.
if inc[i] == None:
grid.append(None)
continue
# Reset.
dist_type = None
end_point = True
# Arccos grid from 0 to pi.
if cdp.params[i] in ['ave_pos_beta', 'eigen_beta', 'axis_theta']:
# Change the default increment numbers.
if not isinstance(inc, list):
inc[i] = int(inc[i] / 2) + 1
# The distribution type and end point.
dist_type = 'acos'
end_point = False
# Append the grid row.
row = grid_row(inc[i], lower[i], upper[i], dist_type=dist_type, end_point=end_point)
grid.append(row)
# Remove an inc if the end point has been removed.
if not end_point:
inc[i] -= 1
# Total number of points.
total_pts = 1
for i in range(n):
# Fixed parameter.
if grid[i] == None:
continue
total_pts = total_pts * len(grid[i])
# Check the number.
max_pts = 50e6
if total_pts > max_pts:
raise RelaxError("The total number of grid points '%s' exceeds the maximum of '%s'." % (total_pts, int(max_pts)))
# Build the points array.
pts = zeros((total_pts, n), float64)
indices = zeros(n, int)
for i in range(total_pts):
# Loop over the dimensions.
for j in range(n):
# Fixed parameter.
if grid[j] == None:
# Get the current parameter value.
pts[i, j] = getattr(cdp, cdp.params[j]) / scaling_matrix[0][j, j]
# Add the point coordinate.
else:
pts[i, j] = grid[j][indices[j]] / scaling_matrix[0][j, j]
# Increment the step positions.
for j in range(n):
if inc[j] != None and indices[j] < inc[j]-1:
indices[j] += 1
break # Exit so that the other step numbers are not incremented.
else:
indices[j] = 0
# Linear constraints.
A, b = None, None
if constraints:
# Obtain the constraints.
A, b = linear_constraints(scaling_matrix=scaling_matrix[0])
# Constraint flag set but no constraints present.
if A is None:
if verbosity:
warn(RelaxWarning("The '%s' model parameters are not constrained, turning the linear constraint algorithm off." % cdp.model))
constraints = False
# The numeric integration information.
if not hasattr(cdp, 'quad_int'):
cdp.quad_int = False
sobol_max_points, sobol_oversample = None, None
if hasattr(cdp, 'sobol_max_points'):
sobol_max_points = cdp.sobol_max_points
sobol_oversample = cdp.sobol_oversample
# Set up the data structures for the target function.
param_vector, full_tensors, full_in_ref_frame, rdcs, rdc_err, rdc_weight, rdc_vect, rdc_const, pcs, pcs_err, pcs_weight, atomic_pos, temp, frq, paramag_centre, com, ave_pos_pivot, pivot, pivot_opt = target_fn_data_setup(sim_index=sim_index, verbosity=verbosity)
# Get the Processor box singleton (it contains the Processor instance) and alias the Processor.
processor_box = Processor_box()
processor = processor_box.processor
# Set up for multi-processor execution.
if processor.processor_size() > 1:
# Printout.
print("Parallelised grid search.")
print("Randomising the grid points to equalise the time required for each grid subdivision.\n")
# Randomise the points.
shuffle(pts)
# Loop over each grid subdivision, with all points violating constraints being eliminated.
for subdivision in grid_split_array(divisions=processor.processor_size(), points=pts, A=A, b=b, verbosity=verbosity):
# Set up the memo for storage on the master.
memo = Frame_order_memo(sim_index=sim_index, scaling_matrix=scaling_matrix[0])
# Set up the command object to send to the slave and execute.
command = Frame_order_grid_command(points=subdivision, scaling_matrix=scaling_matrix[0], sim_index=sim_index, model=cdp.model, param_vector=param_vector, full_tensors=full_tensors, full_in_ref_frame=full_in_ref_frame, rdcs=rdcs, rdc_err=rdc_err, rdc_weight=rdc_weight, rdc_vect=rdc_vect, rdc_const=rdc_const, pcs=pcs, pcs_err=pcs_err, pcs_weight=pcs_weight, atomic_pos=atomic_pos, temp=temp, frq=frq, paramag_centre=paramag_centre, com=com, ave_pos_pivot=ave_pos_pivot, pivot=pivot, pivot_opt=pivot_opt, sobol_max_points=sobol_max_points, sobol_oversample=sobol_oversample, verbosity=verbosity, quad_int=cdp.quad_int)
# Add the slave command and memo to the processor queue.
processor.add_to_queue(command, memo)
# Execute the queued elements.
processor.run_queue()
