def sa_simple(model, map_data, log):
tmp_xrs = model.get_xray_structure().deep_copy_scatterers()
# ro = mmtbx.refinement.real_space.individual_sites.easy(
# map_data = map_data,
# xray_structure = tmp_xrs,
# pdb_hierarchy = model.get_hierarchy().deep_copy(),
# geometry_restraints_manager = model.get_restraints_manager(),
# rms_bonds_limit = 0.01,
# rms_angles_limit = 1.0,
# selection = None, #TODO
# log = log)
# weight = ro.w
weight = 50
#
from mmtbx.dynamics import simulated_annealing as sa
tmp = model.get_xray_structure().deep_copy_scatterers()
params = sa.master_params().extract()
params.start_temperature = 5000
params.cool_rate = 500
sa.run(params=params,
xray_structure=tmp,
real_space=True,
target_map=map_data,
restraints_manager=model.get_restraints_manager(),
wx=weight,
wc=1.,
verbose=False,
log=log)
model.set_sites_cart(sites_cart=tmp.sites_cart())
return model
def sa_simple(rm, xrs, ph, map_data, log):
tmp_xrs = xrs.deep_copy_scatterers()
ro = mmtbx.refinement.real_space.individual_sites.easy(
map_data=map_data,
xray_structure=tmp_xrs,
pdb_hierarchy=ph.deep_copy(),
geometry_restraints_manager=rm,
rms_bonds_limit=0.01,
rms_angles_limit=1.0,
selection=None, #TODO
log=log)
weight = ro.w
#
from mmtbx.dynamics import simulated_annealing as sa
tmp = xrs.deep_copy_scatterers()
params = sa.master_params().extract()
params.start_temperature = 5000
params.cool_rate = 500
sa.run(params=params,
xray_structure=tmp,
real_space=True,
target_map=map_data,
restraints_manager=rm,
wx=weight,
wc=1.,
verbose=False,
log=log)
return tmp.sites_cart()
def run(self, random_seed):
tmp = self.xray_structure.deep_copy_scatterers()
# Shake and minimize to get variable starting points
tmp.shake_sites_in_place(rms_difference=None, mean_distance=5)
sites_cart_ = tmp.sites_cart()
minimized = geometry_minimization.lbfgs(
sites_cart=sites_cart_,
correct_special_position_tolerance=1.0,
geometry_restraints_manager=self.restraints_manager.geometry,
geometry_restraints_flags=self.grf,
lbfgs_exception_handling_params=self.
lbfgs_exception_handling_params,
lbfgs_termination_params=scitbx.lbfgs.termination_parameters(
max_iterations=50))
tmp = tmp.replace_sites_cart(new_sites=sites_cart_)
#
random.seed(random_seed)
flex.set_random_seed(random_seed)
sa.run(params=self.params,
xray_structure=tmp,
real_space=True,
target_map=self.map_data,
restraints_manager=self.restraints_manager,
wx=self.weight,
wc=1.,
verbose=False)
return tmp
def run(self, random_seed):
tmp = self.xray_structure.deep_copy_scatterers()
# Shake and minimize to get variable starting points
tmp.shake_sites_in_place(
rms_difference = None,
mean_distance = 5)
sites_cart_ = tmp.sites_cart()
minimized = geometry_minimization.lbfgs(
sites_cart = sites_cart_,
correct_special_position_tolerance= 1.0,
geometry_restraints_manager = self.restraints_manager.geometry,
geometry_restraints_flags = self.grf,
lbfgs_exception_handling_params = self.lbfgs_exception_handling_params,
lbfgs_termination_params = scitbx.lbfgs.termination_parameters(
max_iterations=50))
tmp = tmp.replace_sites_cart(new_sites = sites_cart_)
#
random.seed(random_seed)
flex.set_random_seed(random_seed)
sa.run(
params = self.params,
xray_structure = tmp,
real_space = True,
target_map = self.map_data,
restraints_manager = self.restraints_manager,
wx = self.weight,
wc = 1.,
verbose = False)
return tmp
def exercise_1():
random.seed(0)
flex.set_random_seed(0)
pi = get_pdb_inputs(pdb_str=pdb_str_1)
f_obs = abs(pi.xrs.structure_factors(d_min=2.5).f_calc())
r_free_flags = f_obs.generate_r_free_flags(use_lattice_symmetry=False)
if (0):
pi.ph.adopt_xray_structure(pi.xrs)
pi.ph.write_pdb_file(file_name="start.pdb",
crystal_symmetry=pi.xrs.crystal_symmetry())
xrs_poor = shake_sites(xrs=pi.xrs.deep_copy_scatterers(),
random=False,
shift=1.5,
grm=pi.grm)
if (0):
pi.ph.adopt_xray_structure(xrs_poor)
pi.ph.write_pdb_file(file_name="poor.pdb",
crystal_symmetry=xrs_poor.crystal_symmetry())
fmodel = mmtbx.f_model.manager(f_obs=f_obs,
r_free_flags=r_free_flags,
xray_structure=xrs_poor)
print("start r_work:", fmodel.r_work())
#
params = sa.master_params().extract()
params.start_temperature = 3000
params.final_temperature = 0
params.cool_rate = 100
params.number_of_steps = 100
params.update_grads_shift = 0.
#
sa.run(params=params,
fmodel=fmodel,
restraints_manager=pi.grm,
wx=20,
wc=1,
verbose=True)
#
r = fmodel.r_work()
print("final r_work:", r)
assert r < 0.03, r
dist = flex.mean(
flex.sqrt(
(pi.xrs.sites_cart() - fmodel.xray_structure.sites_cart()).dot()))
print("Distance(refined, answer): %6.4f" % dist)
assert dist < 0.25, dist
if (0):
pi.ph.adopt_xray_structure(fmodel.xray_structure)
pi.ph.write_pdb_file(
file_name="refined.pdb",
crystal_symmetry=fmodel.xray_structure.crystal_symmetry())
def exercise_1():
random.seed(0)
flex.set_random_seed(0)
pi = get_pdb_inputs(pdb_str=pdb_str_1)
f_obs = abs(pi.xrs.structure_factors(d_min = 2.5).f_calc())
r_free_flags = f_obs.generate_r_free_flags(use_lattice_symmetry=False)
if(0):
pi.ph.adopt_xray_structure(pi.xrs)
pi.ph.write_pdb_file(file_name="start.pdb",
crystal_symmetry = pi.xrs.crystal_symmetry())
xrs_poor = shake_sites(xrs = pi.xrs.deep_copy_scatterers(), random=False,
shift = 1.5, grm=pi.grm)
if(0):
pi.ph.adopt_xray_structure(xrs_poor)
pi.ph.write_pdb_file(file_name="poor.pdb",
crystal_symmetry = xrs_poor.crystal_symmetry())
fmodel = mmtbx.f_model.manager(
f_obs = f_obs,
r_free_flags = r_free_flags,
xray_structure = xrs_poor)
print "start r_work:", fmodel.r_work()
#
params = sa.master_params().extract()
params.start_temperature=3000
params.final_temperature=0
params.cool_rate = 100
params.number_of_steps = 100
params.update_grads_shift = 0.
#
sa.run(
params = params,
fmodel = fmodel,
restraints_manager = pi.grm,
wx = 20,
wc = 1,
verbose = True)
#
r = fmodel.r_work()
print "final r_work:", r
assert r < 0.03, r
dist = flex.mean(flex.sqrt((pi.xrs.sites_cart() -
fmodel.xray_structure.sites_cart()).dot()))
print "Distance(refined, answer): %6.4f"%dist
assert dist < 0.25, dist
if(0):
pi.ph.adopt_xray_structure(fmodel.xray_structure)
pi.ph.write_pdb_file(file_name="refined.pdb",
crystal_symmetry = fmodel.xray_structure.crystal_symmetry())
def anneal(self,
simulated_annealing_params=None,
start_temperature=None,
cool_rate=None,
number_of_steps=50):
"""
Run real-space simulated annealing using the target map (not the RSR map,
if this is different). In practice, the non-selection atoms in the box
should almost always be restrained to their current positions, but the
setup is left to the calling code.
"""
from mmtbx.dynamics import simulated_annealing
import mmtbx.utils
wx = self.real_space_refine(selection=self.selection_all_box)
if (self.debug):
self.box.write_pdb_file("box_start.pdb")
states_collector = None
if (self.debug):
states_collector = mmtbx.utils.states(
xray_structure=self.box.xray_structure_box,
pdb_hierarchy=self.box.pdb_hierarchy_box)
if (simulated_annealing_params is None):
simulated_annealing_params = simulated_annealing.master_params(
).extract()
if (start_temperature is not None):
simulated_annealing_params.start_temperature = start_temperature
if (cool_rate is not None):
simulated_annealing_params.cool_rate = cool_rate
if (number_of_steps is not None):
simulated_annealing_params.number_of_steps = number_of_steps
simulated_annealing.run(params=simulated_annealing_params,
fmodel=None,
xray_structure=self.box.xray_structure_box,
real_space=True,
target_map=self.target_map_box,
restraints_manager=self.box_restraints_manager,
wx=wx,
wc=1.0,
log=self.out,
verbose=True,
states_collector=states_collector)
if (states_collector is not None):
states_collector.write("box_traj.pdb")
self.update_coordinates(self.box.xray_structure_box.sites_cart())
def anneal (self,
simulated_annealing_params=None,
start_temperature=None,
cool_rate=None,
number_of_steps=50) :
"""
Run real-space simulated annealing using the target map (not the RSR map,
if this is different). In practice, the non-selection atoms in the box
should almost always be restrained to their current positions, but the
setup is left to the calling code.
"""
from mmtbx.dynamics import simulated_annealing
import mmtbx.utils
wx = self.real_space_refine(selection=self.selection_all_box)
if (self.debug) :
self.box.write_pdb_file("box_start.pdb")
states_collector = None
if (self.debug) :
states_collector = mmtbx.utils.states(
xray_structure=self.box.xray_structure_box,
pdb_hierarchy=self.box.pdb_hierarchy_box)
if (simulated_annealing_params is None) :
simulated_annealing_params = simulated_annealing.master_params().extract()
if (start_temperature is not None) :
simulated_annealing_params.start_temperature = start_temperature
if (cool_rate is not None) :
simulated_annealing_params.cool_rate = cool_rate
if (number_of_steps is not None) :
simulated_annealing_params.number_of_steps = number_of_steps
simulated_annealing.run(
params = simulated_annealing_params,
fmodel = None,
xray_structure = self.box.xray_structure_box,
real_space = True,
target_map = self.target_map_box,
restraints_manager = self.box_restraints_manager,
wx = wx,
wc = 1.0,
log = self.out,
verbose = True,
states_collector = states_collector)
if (states_collector is not None) :
states_collector.write("box_traj.pdb")
self.update_coordinates(self.box.xray_structure_box.sites_cart())
def exercise_3():
pi = get_pdb_inputs(pdb_str=pdb_str_1)
xrs = pi.xrs.deep_copy_scatterers()
sites_cart_start = xrs.sites_cart()
states_collector = mmtbx.utils.states(pdb_hierarchy=pi.ph)
#
params = sa.master_params().extract()
params.start_temperature = 5000
params.final_temperature = 0
params.cool_rate = 100
params.number_of_steps = 100
params.update_grads_shift = 0.
params.time_step = 0.0005
params.interleave_minimization = True
#
sa.run(params=params,
xray_structure=xrs,
restraints_manager=pi.grm,
states_collector=states_collector)
states_collector.write(file_name="all.pdb")
def exercise_3():
pi = get_pdb_inputs(pdb_str=pdb_str_1)
xrs = pi.xrs.deep_copy_scatterers()
sites_cart_start = xrs.sites_cart()
states_collector = mmtbx.utils.states(
pdb_hierarchy = pi.ph,
xray_structure = xrs)
#
params = sa.master_params().extract()
params.start_temperature=5000
params.final_temperature=0
params.cool_rate = 100
params.number_of_steps = 100
params.update_grads_shift = 0.
params.time_step = 0.0005
params.interleave_minimization=True
#
sa.run(
params = params,
xray_structure = xrs,
restraints_manager = pi.grm,
states_collector = states_collector)
states_collector.write(file_name = "all.pdb")
def exercise_2(d_min = 1.5):
random.seed(2679941)
flex.set_random_seed(2679941)
for shake in [True, False]:
pi = get_pdb_inputs(pdb_str=pdb_str_1)
f_obs = abs(pi.xrs.structure_factors(d_min = d_min).f_calc())
r_free_flags = f_obs.generate_r_free_flags(use_lattice_symmetry=False)
xrs_poor = pi.xrs.deep_copy_scatterers()
if(shake):
xrs_poor = shake_sites(xrs = pi.xrs.deep_copy_scatterers(), random=False,
shift = 2.0, grm=pi.grm)
fmodel = mmtbx.f_model.manager(
f_obs = f_obs,
r_free_flags = r_free_flags,
xray_structure = xrs_poor)
print("start r_work:", fmodel.r_work())
#
f_calc = pi.xrs.structure_factors(d_min = d_min).f_calc()
fft_map = f_calc.fft_map(resolution_factor=0.25)
fft_map.apply_sigma_scaling()
target_map = fft_map.real_map_unpadded()
# find optimal weight
rsr_simple_refiner = mmtbx.refinement.real_space.individual_sites.simple(
target_map = target_map,
selection = flex.bool(pi.xrs.scatterers().size(), True),
real_space_gradients_delta = d_min/4,
max_iterations = 150,
geometry_restraints_manager = pi.grm.geometry)
refined = mmtbx.refinement.real_space.individual_sites.refinery(
refiner = rsr_simple_refiner,
xray_structure = xrs_poor.deep_copy_scatterers(),
start_trial_weight_value = 1,
rms_bonds_limit = 0.02,
rms_angles_limit = 2)
print(refined.weight_final, refined.rms_bonds_final, refined.rms_angles_final)
#
params = sa.master_params().extract()
params.start_temperature=5000
params.final_temperature=0
params.cool_rate = 100
params.number_of_steps = 100
params.update_grads_shift = 0. # does not change runtime visibly
#
sa.run(
params = params,
fmodel = fmodel,
real_space = True,
target_map = target_map,
restraints_manager = pi.grm,
wx = refined.weight_final,
wc = 1.,
verbose = True)
#
r = fmodel.r_work()
print("final r_work:", r)
if(shake):
assert r < 0.07, r
else:
assert r < 0.07, r
dist = flex.mean(flex.sqrt((pi.xrs.sites_cart() -
fmodel.xray_structure.sites_cart()).dot()))
print("Distance(refined, answer): %6.4f"%dist)
if(shake):
assert dist < 0.35, r
else:
assert dist < 0.06, r
if(0):
pi.ph.adopt_xray_structure(fmodel.xray_structure)
pi.ph.write_pdb_file(file_name="refined.pdb",
crystal_symmetry = fmodel.xray_structure.crystal_symmetry())
def run(self):
hierarchy = self.model.get_hierarchy()
map_data, grid_unit_cell = None, None
#### <Begin> sanity check for map and model
if self.map_inp is not None:
base = map_and_model.input(map_data=self.map_inp.map_data(),
model=self.model,
crystal_symmetry=self.cs_consensus,
box=False)
hierarchy = base.model().get_hierarchy()
map_data = base.map_data()
grid_unit_cell = self.map_inp.grid_unit_cell()
hierarchy.atoms().reset_i_seq()
#### <End> sanity check for map and model
# Initialize states accumulator # Pavel's original
states = mmtbx.utils.states(
pdb_hierarchy=self.model.get_hierarchy(),
xray_structure=self.model.get_xray_structure())
states.add(sites_cart=self.model.get_xray_structure().sites_cart())
params = sa.master_params().extract(
) # because of params = sa.master_params().extract() above, core parameters need to be redefined
params.start_temperature = self.params.start_temperature
params.final_temperature = self.params.final_temperature
params.cool_rate = self.params.cool_rate
#params.MD_in_each_cycle = self.params.MD_in_each_cycle # "AttributeError: Assignment to non-existing attribute "MD_in_each_cycle"
params.number_of_steps = self.params.number_of_steps
max_steps_for_final_MD = ''
if (self.params.max_steps_for_final_MD != None):
max_steps_for_final_MD = self.params.max_steps_for_final_MD
params.update_grads_shift = 0.
params.interleave_minimization = False #Pavel will fix the error that occur when params.interleave_minimization=True
#print ("params:",params) # object like <libtbx.phil.scope_extract object at 0x1146ae210>
map_inp = self.map_inp
user_map_weight = self.user_map_weight
map_weight_multiply = self.map_weight_multiply
if (
self.params.record_states == False
): # default choice to avoid > 160 GB memory issue with recording all states for L1 stalk
states = None
if (self.params.reoptimize_map_weight_after_each_cycle_during_final_MD
== True):
cycle_so_far_for_map_weight_reoptimization = 0
splited_model_name = self.model_name[:-4].split("/")
model_file_name_only = splited_model_name[len(splited_model_name) - 1]
#number_of_atoms_in_input_pdb = know_number_of_atoms_in_input_pdb(self.logfile, self.model_name)
# tRNA : 1,563
# L1 stalk : 3,289
# Mg channel: 14,940
# number_of_atoms_in_input_pdb seems irrelevant to check_cc_after_these_cycles assignment.
# but Mg channel with 10k check took 10 days!
#### <begin> prepare/initialize for iteration
check_cc_after_these_steps = '' # use '# of steps' not '# of iterations'
if (("tst_cryo_fit2" in model_file_name_only) == True):
#check_cc_after_these_steps = 1000 # tst_2 took 2 min?
check_cc_after_these_steps = 700
# if this is too small (like 100), it may run forever
# I confirmed that 500 is definitely too small to explore properly (a helix),
#but this is just for test
else:
check_cc_after_these_steps = 10000
#check_cc_after_these_steps = 2000 #even if this value is so small like this, empty 1st_2nd_array error is avoided by following fail-proof hook
number_of_MD_in_each_cycle = 1 + (
(params.start_temperature - params.final_temperature) /
params.cool_rate)
# same value as MD_in_each_cycle
# Regardless of above assignment, re-assign check_cc_after_these_steps to avoid empty 1st_2nd_array situation
check_cc_after_these_steps = check_cc_after_these_steps + params.number_of_steps * number_of_MD_in_each_cycle * 2
# reoptimize_map_weight_after_these_steps = ''
# if (self.params.reoptimize_map_weight_after_each_cycle_during_final_MD == True):
# if (("tst_cryo_fit2" in model_file_name_only) == True):
# reoptimize_map_weight_after_these_steps = 5
# else:
# reoptimize_map_weight_after_these_steps = 100 # after 123~171 cycles, full tRNA crashes (when map_weight is multiplied too crazy back then,,,)
if (("tst_cryo_fit2_" in self.model_name) == True):
self.params.max_steps_for_exploration = 100
#max_steps_for_final_MD = 10000
#max_steps_for_final_MD = 3000
map_weight_before_multiplication = self.params.map_weight
self.params.map_weight = self.params.map_weight * map_weight_multiply
#### This is the only place where map_weight_multiply is applied (other than reoptimize_map_weight_if_not_specified for final MD)
best_cc_so_far = -999 # tRNA has a negative value of initial cc
cc_1st_array = []
cc_2nd_array = []
result = ''
total_steps_so_far_for_exploration_and_final_MD = 0
total_steps_so_far_for_cc_check = 0 # initialization
#### <end> prepare/initialize for iteration
grm = self.model.get_restraints_manager()
pdb_hierarchy = self.model.get_hierarchy()
'''
#get_stacking_proxies() takes exactly 5 arguments
stacking_proxies = nucleic_acids.get_stacking_proxies(
pdb_hierarchy = pdb_hierarchy,
stacking_phil_params = self.params.secondary_structure.nucleic_acid.stacking_pair,
grm=grm) #,
#mon_lib_srv=self.mon_lib_srv, # AttributeError: 'cryo_fit2_class' object has no attribute 'mon_lib_sr
#plane_cache=plane_cache)
print(" %d stacking parallelities" % len(stacking_proxies), file=log)
STOP()
'''
write_this = "\nself.params.map_weight after multiplication (" + str(
map_weight_multiply) + ") = " + str(
round(self.params.map_weight, 1)) + "\n"
print(write_this)
self.logfile.write(str(write_this))
########################### <begin> iterate until cryo_fit2 derived cc saturates
for i in range(
100000000
): # runs well with cryo_fit2.run_tests #for i in range(1000000000): # fails with cryo_fit2.run_tests with too much memory (bigger than 30 GB)
write_this = "\n" + str(i + 1) + "th iteration: \n"
print(write_this)
self.logfile.write(str(write_this))
try:
if (self.params.progress_on_screen == True): # default choice
result = sa.run(
params=params,
xray_structure=self.model.get_xray_structure(),
#restraints_manager = self.model.get_restraints_manager(),
restraints_manager=grm,
target_map=map_data,
real_space=True,
wx=self.params.map_weight,
wc=1, # weight for geometry conformation
states_collector=states)
else: # (self.params.progress_on_screen = False):
result = sa.run(
params=params,
xray_structure=self.model.get_xray_structure(),
#restraints_manager = self.model.get_restraints_manager(),
restraints_manager=grm,
target_map=map_data,
real_space=True,
wx=self.params.map_weight,
wc=1, # weight for geometry conformation
states_collector=states,
log=self.logfile
) # if this is commented, temp= xx dist_moved= xx angles= xx bonds= xx is shown on screen rather than cryo_fit2.log
except Exception as ex:
write_this = "exception message:" + str(ex)
print(write_this)
self.logfile.write(str(write_this))
write_this = "Failed during core map weight multiplied phenix.dynamics run.\n"
print(write_this)
self.logfile.write(str(write_this))
return self.output_dir
total_steps_so_far_for_exploration_and_final_MD = total_steps_so_far_for_exploration_and_final_MD \
+ int(params.number_of_steps*number_of_MD_in_each_cycle)
cc_after_small_MD = calculate_overall_cc(
map_data=map_data,
model=self.model,
resolution=self.params.resolution)
write_this = "CC after this cycle (a small MD iteration): " + str(
round(cc_after_small_MD, 7)) + "\n"
self.logfile.write(str(write_this))
if (self.params.explore == True):
if (total_steps_so_far_for_exploration_and_final_MD <
self.params.max_steps_for_exploration):
write_this = "\ntotal_steps_so_far_for_exploration_and_final_MD (" + str(total_steps_so_far_for_exploration_and_final_MD) + \
") < max_steps_for_exploration (" + str(self.params.max_steps_for_exploration) + ")\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
continue
else:
write_this = "\ntotal_steps_so_far_for_exploration_and_final_MD (" + str(total_steps_so_far_for_exploration_and_final_MD) + \
") >= max_steps_for_exploration (" + str(self.params.max_steps_for_exploration) + ")\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
break
############# All below is for final MD
total_steps_so_far_for_cc_check = total_steps_so_far_for_cc_check + int(
params.number_of_steps * number_of_MD_in_each_cycle)
cc_improvement_threshold = ''
if (("tst_cryo_fit2" in model_file_name_only) == True):
#cc_improvement_threshold = 0.01 # to finish regression quickly # took 2 min for tst_2?
cc_improvement_threshold = 0.05 # to finish regression quickly # took 2 min for tst_2?
else:
cc_improvement_threshold = 0.00001 # even a 0.0001 improved cc further eventually significantly
if (max_steps_for_final_MD != ''):
if (total_steps_so_far_for_exploration_and_final_MD >=
max_steps_for_final_MD):
write_this = ''
if (self.params.explore == True):
write_this = "\ntotal_steps_so_far_for_exploration_and_final_MD (" + str(total_steps_so_far_for_exploration_and_final_MD) + \
") >= A specified max_steps_for_final_MD (" + str(max_steps_for_final_MD) + ")\n"
else:
write_this = "\ntotal steps final_MD (" + str(total_steps_so_far_for_exploration_and_final_MD) + \
") >= A specified max_steps_for_final_MD (" + str(max_steps_for_final_MD) + ")\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
break
if (float(total_steps_so_far_for_cc_check) < float(
check_cc_after_these_steps / 2)):
cc_1st_array.append(cc_after_small_MD)
else:
cc_2nd_array.append(cc_after_small_MD)
'''
if (self.params.reoptimize_map_weight_after_each_cycle_during_final_MD == True):
if (cycle_so_far_for_map_weight_reoptimization >= reoptimize_map_weight_after_these_steps):
self.params.map_weight = reoptimize_map_weight_if_not_specified(self, user_map_weight, map_inp)
self.params.map_weight = self.params.map_weight * map_weight_multiply
cycle_so_far_for_map_weight_reoptimization = 0 # reinitialization
# I confirmed that reoptimizing map_weight_after_each_cycle did change result (cc, SS stat) significantly
'''
# total_steps_so_far_for_cc_check is thought to be re-initialized in all circumstances. However, it seems not.
if (total_steps_so_far_for_cc_check >= check_cc_after_these_steps):
if (cc_after_small_MD > best_cc_so_far):
write_this = "current_cc (" + str(
cc_after_small_MD) + ") > best_cc_so_far (" + str(
best_cc_so_far
) + "). \nTherefore, cryo_fit2 will run longer MD.\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
write_this = "cc_after_small_MD - best_cc_so_far = " + str(
float_to_str(cc_after_small_MD - best_cc_so_far)
) + "\n" # this "\n" is essential for screen print
print('%s' % (write_this))
self.logfile.write(str(write_this))
if (
float(cc_after_small_MD - best_cc_so_far) >
cc_improvement_threshold
): # without this if clause, later MD cycles that improve just tiny fractions of cc take too long time
write_this = "cc_after_small_MD - best_cc_so_far > cc_improvement_threshold (" + str(
float_to_str(cc_improvement_threshold)
) + "). Iterates longer.\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
best_cc_so_far = cc_after_small_MD
cc_1st_array = [] # reset
cc_2nd_array = [] # reset
total_steps_so_far_for_cc_check = 0 # reset
continue
else:
write_this = "cc_after_small_MD - best_cc_so_far <= " + str(
float_to_str(cc_improvement_threshold)
) + ". Goes to mean_array_comparison.\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
else:
write_this = "current_cc (" + str(
cc_after_small_MD) + ") <= best_cc_so_far (" + str(
best_cc_so_far) + ")\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
if (np.mean(cc_2nd_array) > np.mean(cc_1st_array)):
write_this = "mean of cc_2nd_array (" + str(
np.mean(cc_2nd_array)
) + ") > mean of cc_1st_array (" + str(
np.mean(cc_1st_array)) + ")\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
write_this = "(mean of cc_2nd_array) - (mean of cc_1st_array): " + str(
np.mean(cc_2nd_array) - np.mean(cc_1st_array)) + "\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
if (
(np.mean(cc_2nd_array) - np.mean(cc_1st_array)) >
cc_improvement_threshold
): # without this if clause, later MD cycles that improve just tiny fractions of cc take too long time
cc_1st_array = [] # reset
cc_2nd_array = [] # reset
total_steps_so_far_for_cc_check = 0 # reset
else:
write_this = "cc values are saturated\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
if (self.params.explore == True
): # no need to report cc after explore
write_this = "total_steps_so_far_for_exploration_and_final_MD: " + str(
total_steps_so_far_for_exploration_and_final_MD
) + "\n"
else: # (self.params.explore = False): # no need to report cc after explore
write_this = "total_steps for final_MD: " + str(
total_steps_so_far_for_exploration_and_final_MD
) + "\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
break
else: #(np.mean(cc_2nd_array) <= np.mean(cc_1st_array)):
write_this = "mean of cc_2nd_array (" + str(
np.mean(cc_2nd_array)
) + ") <= mean of cc_1st_array (" + str(
np.mean(cc_1st_array)) + ")\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
write_this = "cc values are saturated\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
if (self.params.explore == True
): # no need to report cc after explore
write_this = "total_steps_so_far_for_exploration_and_final_MD: " + str(
total_steps_so_far_for_exploration_and_final_MD
) + "\n"
else: # (self.params.explore = False): # no need to report cc after explore
write_this = "total_steps for final_MD: " + str(
total_steps_so_far_for_exploration_and_final_MD
) + "\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
break
######################### <end> iterate until cryo_fit2 derived cc saturates
overall_cc_after_cryo_fit2 = calculate_overall_cc(
map_data=map_data,
model=self.model,
resolution=self.params.resolution)
write_this = "\nFinal MD of cryo_fit2 is done.\n"
print('%s' % (write_this))
self.logfile.write(str(write_this))
if (self.params.explore == False
): # no need to report cc after explore
''' since this differs from CC_box, let's not use this (??)
write_this = "\nCC_overall after cryo_fit2 (both exploration and final MD): " + str(round(overall_cc_after_cryo_fit2, 4)) + "\n"
print('%s' %(write_this))
self.logfile.write(str(write_this))
#'''
output_dir_w_CC = str(self.output_dir) + "_cc_" + str(
round(overall_cc_after_cryo_fit2, 3))
if os.path.exists(output_dir_w_CC):
shutil.rmtree(output_dir_w_CC)
os.mkdir(output_dir_w_CC)
if (self.params.record_states == True):
all_state_file = os.path.join(output_dir_w_CC, "all_states.pdb")
states.write(file_name=all_state_file)
self.model.set_xray_structure(result.xray_structure)
fitted_file_name = model_file_name_only + "_cryo_fit2_fitted.pdb"
fitted_file_name_w_path = os.path.join(output_dir_w_CC,
fitted_file_name)
##### this is essential to spit cryo_fitted2 file
with open(fitted_file_name_w_path, "w") as f:
f.write(self.model.model_as_pdb())
f.close()
#print_this ='''
######## How to fix map origin problem in cryo_fit2 #######
'''
With 0,0,0 origin map, cryo_fit2 has no problem.
However, with non-0,0,0 origin cryo-EM map, cryo_fit2 results cryo_fitted pdb model at "wrong" origin
This is because probably dynamics part uses map at 0,0,0 origin.
Therefore, cryo_fit2 identifies how much the map origin was moved, then update all xyz coordinates of output pdb file.
In user's perspective, there is nothing to bother.
All kinds of mrc files (e.g. "Regular", emdb downloaded, went through phenix.map_box, gaussian filtered by UCSF Chimera and went through relion_image_handler) work fine.
#############################################################
#print (print_this,"\n")
'''
try:
bp_num_in_fitted_file, sp_num_in_fitted_file, H_num_in_fitted_file, E_num_in_fitted_file = \
count_bp_sp_H_E_in_fitted_file(fitted_file_name_w_path, output_dir_w_CC, self.logfile)
except Exception as ex:
write_this = "exception message:" + str(ex)
print(write_this)
self.logfile.write(str(write_this))
write_this = "(in task_obj loop) An exception occurred in cryo_fit2_run. \n" + \
" Maybe cryo_fit2 failed to run (\"nan\" or secondary_structure_restraint file generataion failure) for this condition:" + \
" cool_rate (" + str(round(params.cool_rate, 1)) + ")\n" + \
" number_of_steps (" + str(params.number_of_steps) + ")\n" + \
" start_temperature (" + str(params.start_temperature) + ")\n" + \
" map_weight_multiply (" + str(map_weight_multiply) + ")\n" + \
" final_temperature (" + str(params.final_temperature) + ")\n" + \
" map_weight (" + str(round(self.params.map_weight,2)) + ")\n" + \
" max_steps_for_final_MD (" + str(max_steps_for_final_MD) + ")"
print(write_this)
self.logfile.write(str(write_this))
if (os.path.isdir("parameters_exploration/bp_H_E_not_calculated")
== False):
os.mkdir("parameters_exploration/bp_H_E_not_calculated")
command_string = "mv " + str(
output_dir_w_CC
) + " parameters_exploration/bp_H_E_not_calculated"
logfile.write(str(command_string))
libtbx.easy_run.fully_buffered(
command=command_string).raise_if_errors().stdout_lines
return output_dir_w_CC
returned = know_how_much_map_origin_moved(str(self.map_name))
if (returned != "origin_is_all_zero"
and self.params.keep_origin == True):
write_this = "Restoring original xyz position for a cryo_fit2 fitted atomistic model\n"
print(write_this)
self.logfile.write(str(write_this))
return_to_origin_of_pdb_file(fitted_file_name_w_path, returned[0],
returned[1], returned[2], returned[3])
if (("tst_cryo_fit2" in fitted_file_name_w_path) == False):
calculate_RMSD(self, fitted_file_name_w_path)
output_dir_final = output_dir_w_CC + "_bp_" + str(bp_num_in_fitted_file) + "_sp_" + str(sp_num_in_fitted_file) \
+ "_H_" + str(H_num_in_fitted_file) + "_E_" + str(E_num_in_fitted_file)
if os.path.exists(output_dir_final):
shutil.rmtree(output_dir_final)
mv_command_string = "mv " + output_dir_w_CC + " " + output_dir_final
libtbx.easy_run.fully_buffered(mv_command_string)
############################
current_dir = os.getcwd()
os.chdir(output_dir_final)
command_string = "echo " + str(
map_weight_before_multiplication
) + " >> used_map_weight_before_multiplication.txt"
libtbx.easy_run.fully_buffered(
command=command_string).raise_if_errors().stdout_lines
os.chdir(current_dir)
############################
return output_dir_final
def run(args, prefix="tst_00", validated=False):
user_input_pdb = ''
user_input_map = ''
# very simple parsing of model and map
for i, arg in enumerate(args):
if arg.endswith('.cif') or arg.endswith('.ent') or arg.endswith('.pdb'): # EMD-3981 has 6exv.ent instead of .pdb
user_input_pdb = arg
if arg.find('=')==-1:
args[i]='model=%s' % arg
elif arg.endswith('.ccp4') or arg.endswith('.map'):
user_input_map = arg
if arg.find('=')==-1:
args[i]='map=%s' % arg
argument_interpreter = libtbx.phil.command_line.argument_interpreter(
master_phil=master_phil,
home_scope="cryo_fit2",
)
user_input_pdb = clean_pdb_for_phenix(user_input_pdb)
pdbs = []
maps = []
phils = []
phil_args = []
for arg in args:
if os.path.isfile(arg) :
if iotbx.pdb.is_pdb_file(arg):
pdbs.append(arg)
elif arg.endswith('.ccp4') or arg.endswith('.map'): # not the smartest
maps.append(arg)
else:
try :
file_phil = phil.parse(file_name=arg)
except RuntimeError :
pass
else :
phils.append(file_phil)
else :
phil_args.append(arg)
phils.append(argument_interpreter.process(arg))
working_phil = master_phil.fetch(sources=phils)
working_phil.show()
working_params = working_phil.extract()
if (not validated):
validate_params(working_params)
# Compute a target map
from iotbx import ccp4_map
ccp4_map = ccp4_map.map_reader(user_input_map)
print('Map read from %s' %(user_input_map))
target_map_data = ccp4_map.map_data()
# initial atomic model that we want to fit to an EM-map
pdb_inp = iotbx.pdb.input(file_name=user_input_pdb)
model = mmtbx.model.manager(model_input = pdb_inp)
# Initialize states accumulator
states = mmtbx.utils.states(
pdb_hierarchy = model.get_hierarchy(),
xray_structure = model.get_xray_structure())
states.add(sites_cart = model.get_xray_structure().sites_cart())
#
params = sa.master_params().extract()
params.start_temperature=2000
params.final_temperature=0
params.cool_rate = 100
params.number_of_steps = 1000
params.update_grads_shift = 0.
params.interleave_minimization=False #Pavel will fix the error that occur when params.interleave_minimization=True
print('CC: %s' %(calculate_cc(map_data=target_map_data, model=model, resolution=3.)))
#STOP()
result = sa.run(
params = params,
xray_structure = model.get_xray_structure(),
restraints_manager = model.get_restraints_manager(),
target_map = target_map_data,
real_space = True,
wx = 100, # wx=5 broke helix conformation of tst_00_poor.pdb, wx=100 kept helix well
wc = 1,
states_collector = states)
states.write(file_name = "all_states.pdb")
model.set_xray_structure(result.xray_structure)
with open("refined.pdb", "w") as f:
f.write(model.model_as_pdb())
def exercise_2(d_min = 1.5):
random.seed(2679941)
flex.set_random_seed(2679941)
for shake in [True, False]:
pi = get_pdb_inputs(pdb_str=pdb_str_1)
f_obs = abs(pi.xrs.structure_factors(d_min = d_min).f_calc())
r_free_flags = f_obs.generate_r_free_flags(use_lattice_symmetry=False)
xrs_poor = pi.xrs.deep_copy_scatterers()
if(shake):
xrs_poor = shake_sites(xrs = pi.xrs.deep_copy_scatterers(), random=False,
shift = 2.0, grm=pi.grm)
fmodel = mmtbx.f_model.manager(
f_obs = f_obs,
r_free_flags = r_free_flags,
xray_structure = xrs_poor)
print "start r_work:", fmodel.r_work()
#
f_calc = pi.xrs.structure_factors(d_min = d_min).f_calc()
fft_map = f_calc.fft_map(resolution_factor=0.25)
fft_map.apply_sigma_scaling()
target_map = fft_map.real_map_unpadded()
# find optimal weight
rsr_simple_refiner = mmtbx.refinement.real_space.individual_sites.simple(
target_map = target_map,
selection = flex.bool(pi.xrs.scatterers().size(), True),
real_space_gradients_delta = d_min/4,
max_iterations = 150,
geometry_restraints_manager = pi.grm.geometry)
refined = mmtbx.refinement.real_space.individual_sites.refinery(
refiner = rsr_simple_refiner,
xray_structure = xrs_poor.deep_copy_scatterers(),
start_trial_weight_value = 1,
rms_bonds_limit = 0.02,
rms_angles_limit = 2)
print refined.weight_final, refined.rms_bonds_final, refined.rms_angles_final
#
params = sa.master_params().extract()
params.start_temperature=5000
params.final_temperature=0
params.cool_rate = 100
params.number_of_steps = 100
params.update_grads_shift = 0. # does not change runtime visibly
#
sa.run(
params = params,
fmodel = fmodel,
real_space = True,
target_map = target_map,
restraints_manager = pi.grm,
wx = refined.weight_final,
wc = 1.,
verbose = True)
#
r = fmodel.r_work()
print "final r_work:", r
if(shake):
assert r < 0.07, r
else:
assert r < 0.07, r
dist = flex.mean(flex.sqrt((pi.xrs.sites_cart() -
fmodel.xray_structure.sites_cart()).dot()))
print "Distance(refined, answer): %6.4f"%dist
if(shake):
assert dist < 0.35, r
else:
assert dist < 0.06, r
if(0):
pi.ph.adopt_xray_structure(fmodel.xray_structure)
pi.ph.write_pdb_file(file_name="refined.pdb",
crystal_symmetry = fmodel.xray_structure.crystal_symmetry())
