def changeLoop(filename, loopname):
posted=request.get_json(force=True)
if posted["action"]=="change" and posted["method"]=="random":
sm=get_sm(filename)
sm.load_sampled_elems()
if loopname not in sm.bg.coords:
abort(404)
original_cg=copy.deepcopy(sm.bg)
new_filename=get_new_filename()
smCache.renameSM(filename, new_filename)
change_elem(sm, loopname)
cg=sm.bg
centroid0 = ftuv.get_vector_centroid(ftug.bg_virtual_residues(original_cg))
centroid1 = ftuv.get_vector_centroid(ftug.bg_virtual_residues(cg))
crds0 = ftuv.center_on_centroid(ftug.bg_virtual_residues(original_cg))
crds1 = ftuv.center_on_centroid(ftug.bg_virtual_residues(cg))
rot_mat = ftur.optimal_superposition(crds0, crds1)
for k in cg.coords.keys():
cg.coords[k] = (np.dot(rot_mat, cg.coords[k][0] - centroid1),
np.dot(rot_mat, cg.coords[k][1] - centroid1))
if k[0] == 's':
cg.twists[k] = (np.dot(rot_mat, cg.twists[k][0]),
np.dot(rot_mat, cg.twists[k][1]))
filename=get_new_filename()
cg.to_cg_file("user_files/"+filename)
return jsonify({"url": url_for("structure_main", filename=filename)})
else:
abort(403)
def test_cg_rmsd(self):
cg1 = ftmc.CoarseGrainRNA('test/forgi/threedee/data/1GID_A.cg')
cg2 = ftmc.CoarseGrainRNA('test/forgi/threedee/data/1GID_A_sampled.cg')
residues1 = ftug.bg_virtual_residues(cg1)
residues2 = ftug.bg_virtual_residues(cg2)
self.assertAlmostEqual(ftme.rmsd(residues1, residues2),
ftme.cg_rmsd(cg1, cg2))
def main():
usage = """
python cg_rmsd.py file1.cg file2.cg
Calculate the RMSD between two coarse grain models.
"""
num_args= 2
parser = OptionParser(usage=usage)
#parser.add_option('-o', '--options', dest='some_option', default='yo', help="Place holder for a real option", type='str')
#parser.add_option('-u', '--useless', dest='uselesss', default=False, action='store_true', help='Another useless option')
(options, args) = parser.parse_args()
if len(args) < num_args:
parser.print_help()
sys.exit(1)
cg1 = ftmc.CoarseGrainRNA(args[0])
cg2 = ftmc.CoarseGrainRNA(args[1])
vrs1 = ftug.bg_virtual_residues(cg1)
vrs2 = ftug.bg_virtual_residues(cg2)
print ftur.centered_rmsd(vrs1, vrs2)
def changeLoop(filename, loopname):
posted = request.get_json(force=True)
if posted["action"] == "change" and posted["method"] == "random":
sm = get_sm(filename)
sm.load_sampled_elems()
if loopname not in sm.bg.coords:
abort(404)
original_cg = copy.deepcopy(sm.bg)
new_filename = get_new_filename()
smCache.renameSM(filename, new_filename)
change_elem(sm, loopname)
cg = sm.bg
centroid0 = ftuv.get_vector_centroid(
ftug.bg_virtual_residues(original_cg))
centroid1 = ftuv.get_vector_centroid(ftug.bg_virtual_residues(cg))
crds0 = ftuv.center_on_centroid(ftug.bg_virtual_residues(original_cg))
crds1 = ftuv.center_on_centroid(ftug.bg_virtual_residues(cg))
rot_mat = ftur.optimal_superposition(crds0, crds1)
for k in cg.coords.keys():
cg.coords[k] = (np.dot(rot_mat, cg.coords[k][0] - centroid1),
np.dot(rot_mat, cg.coords[k][1] - centroid1))
if k[0] == 's':
cg.twists[k] = (np.dot(rot_mat, cg.twists[k][0]),
np.dot(rot_mat, cg.twists[k][1]))
filename = get_new_filename()
cg.to_cg_file("user_files/" + filename)
return jsonify({"url": url_for("structure_main", filename=filename)})
else:
abort(403)
def align_cgs(cgs):
'''
Align each coarse grain RNA to the first one.
The points representing each coarse grain RNA molecule
will be the virtual residues.
@param cgs: A list of CoarseGrainRNA structures.
@return: Nothing, the cgs are modified in place
'''
centroid0 = ftuv.get_vector_centroid(ftug.bg_virtual_residues(cgs[0]))
crds0 = ftuv.center_on_centroid(ftug.bg_virtual_residues(cgs[0]))
for cg in cgs:
centroid1 = ftuv.get_vector_centroid(ftug.bg_virtual_residues(cg))
crds1 = ftuv.center_on_centroid(ftug.bg_virtual_residues(cg))
rot_mat = ftur.optimal_superposition(crds0, crds1)
for k in cg.coords.keys():
cg.coords[k] = (np.dot(rot_mat, cg.coords[k][0] - centroid1),
np.dot(rot_mat, cg.coords[k][1] - centroid1))
if k[0] == 's':
cg.twists[k] = (np.dot(rot_mat, cg.twists[k][0]),
np.dot(rot_mat, cg.twists[k][1]))
def __init__(self, sm_orig, plotter=None, plot_color=None, silent=False, output_file=sys.stdout, save_n_best=3, dists=[], no_rmsd=False, save_iterative_cg_measures=False):
'''
@param sm_orig: The original Spatial Model against which to collect statistics.
'''
self.energy_rmsd_structs = []
self.counter = 0
self.plotter = plotter
self.plot_color = plot_color
self.silent = silent
self.verbose = False
self.output_file = output_file
self.save_n_best = save_n_best
self.sm_orig = sm_orig
self.energy_orig = None
self.step_save = 0
self.save_iterative_cg_measures = save_iterative_cg_measures
self.dists = dists
self.highest_rmsd = 0.
self.lowest_rmsd = 10000000000.
self.no_rmsd = no_rmsd
self.creation_time = time.time()
try:
self.centers_orig = ftug.bg_virtual_residues(sm_orig.bg)
self.confusion_matrix_calculator = ftme.ConfusionMatrix(sm_orig.bg)
except KeyError:
# if there are no coordinates provided in the original
# bulge graph file, then don't calculate rmsds
self.centers_orig = None
self.confusion_matrix_calculator = None
def main(args):
cgs = []
projs = []
for file_ in args.files:
cgs.append(ftmc.CoarseGrainRNA(file_))
try:
projs.append(ftmp.Projection2D(cgs[-1]))
except ValueError:
projs.append(None)
p_rmsds = OrderedDict()
cg_rmsds = OrderedDict()
p_rmsds[0] = 0.0
cg_rmsds[0] = 0.0
if "," in args.max_diff:
diffs = map(int, args.max_diff.split(","))
else:
diffs = range(1, int(args.max_diff) + 1)
for diff in diffs:
print ("diff {}".format(diff))
prmsd = 0
cgrmsd = 0
count = 0
pcount = 0
for i in range(0, len(cgs) - diff):
count += 1
try:
vrs1 = np.array([x for p in sorted(projs[i]._coords.keys()) for x in projs[i]._coords[p]])
vrs2 = np.array([x for p in sorted(projs[i + diff]._coords.keys()) for x in projs[i + diff]._coords[p]])
prmsd += ftms.rmsd(vrs1, vrs2)
pcount += 1
except:
pass
cgrmsd += ftms.cg_rmsd(cgs[i], cgs[i + diff])
if count:
cg_rmsds[diff] = cgrmsd / count
if pcount:
p_rmsds[diff] = prmsd / pcount
print "projection RMSDs:", p_rmsds
print "3D-RMSDs:", cg_rmsds
import matplotlib.pyplot as plt
if args.target_structure and args.hausdorff_reference:
fig, (ax, axT, axH) = plt.subplots(3)
elif args.target_structure:
fig, (ax, axT) = plt.subplots(2)
elif args.hausdorff_reference:
fig, (ax, axH) = plt.subplots(2)
else:
fig, ax = plt.subplots()
if args.target_structure:
target = ftmc.CoarseGrainRNA(args.target_structure)
target_rmsds = []
target_proj_rmsds = []
try:
target_proj = ftmp.Projection2D(target)
except ValueError:
pass
else:
target_vrs = np.array([x for p in sorted(target_proj._coords.keys()) for x in target_proj._coords[p]])
for proj in projs:
try:
vrs1 = np.array([x for p in sorted(proj._coords.keys()) for x in proj._coords[p]])
prmsd = ftms.rmsd(vrs1, target_vrs)
target_proj_rmsds.append(prmsd)
except:
target_proj_rmsds.append(float("nan"))
target_vrs = ftug.bg_virtual_residues(target)
for cg in cgs:
vrs1 = ftug.bg_virtual_residues(cg)
cgrmsd = ftms.rmsd(vrs1, target_vrs)
target_rmsds.append(cgrmsd)
xval = np.arange(len(cgs)) * args.step_size
if target_proj_rmsds:
axT.plot(xval, target_proj_rmsds, label="Projection", color="green")
axT.plot(xval, target_rmsds, label="3D", color="blue")
axT.set_title("RMSD to target structure")
axT.set_xlabel("step")
axT.set_ylabel("RMSD")
leg = axT.legend(framealpha=0.8, fancybox=True)
leg.get_frame().set_linewidth(0.0)
plt.subplots_adjust(hspace=0.4)
if args.hausdorff_reference:
ref_img = scipy.ndimage.imread(args.hausdorff_reference)
ref_quarter = scipy.ndimage.zoom(ref_img, 0.3) > 150
degrees = get_refimg_longest_axis(ref_img)
global_optima = []
local_optima = []
for cg in cgs:
score, img, params = fph.try_parameters(ref_img, args.hausdorff_scale, cg, degrees)
local_optima.append(score)
s, i, params = fph.globally_minimal_distance(
ref_quarter, args.hausdorff_scale, cg, start_points=40, starting_rotations=degrees, virtual_atoms=False
)
score, img, params = fph.locally_minimal_distance(
ref_img, args.hausdorff_scale, cg, params[1], params[2], params[0], maxiter=200
)
global_optima.append(score)
axH.plot(xval, global_optima, label="Global Minimization", color="red")
axH.plot(xval, local_optima, label="Local Minimization", color="lavender")
axH.set_title("Hausdorff Distance to reference image")
axH.set_xlabel("step")
axH.set_ylabel("Hausdorff Distance")
leg = axH.legend(framealpha=0.8, fancybox=True)
leg.get_frame().set_linewidth(0.0)
plt.subplots_adjust(hspace=0.4)
p_rmsds = np.array(p_rmsds.items())
cg_rmsds = np.array(cg_rmsds.items())
ax.plot(p_rmsds[:, 0] * args.step_size, p_rmsds[:, 1], label="Projection", color="green", marker="o")
ax.plot(cg_rmsds[:, 0] * args.step_size, cg_rmsds[:, 1], label="3D", color="blue", marker="o")
ax.set_title("Average RMSD between structures X steps apart.")
ax.set_xlabel("steps apart")
ax.set_ylabel("RMSD")
leg = ax.legend(framealpha=0.8, fancybox=True)
leg.get_frame().set_linewidth(0.0)
fig.patch.set_facecolor("white")
if args.save_fig:
with open(args.save_fig, "w") as f:
pickle.dump(fig, f)
plt.show()
def test_cg_rmsd(self):
cg1 = ftmc.CoarseGrainRNA('test/forgi/threedee/data/1GID_A.cg')
cg2 = ftmc.CoarseGrainRNA('test/forgi/threedee/data/1GID_A_sampled.cg')
residues1 = ftug.bg_virtual_residues(cg1)
residues2 = ftug.bg_virtual_residues(cg2)
self.assertAlmostEqual(ftur.centered_rmsd(residues1, residues2), ftme.cg_rmsd(cg1,cg2))
#! /usr/bin/python
import forgi.threedee.model.coarse_grain as ftmc
import fess.builder.models as fbm
import forgi.threedee.utilities.vector as ftuv
import forgi.threedee.utilities.graph_pdb as ftug
import sys
if __name__=="__main__":
filename=sys.argv[1]
print ("BUILDSTRUCTURE: Filename is ", filename)
cg=ftmc.CoarseGrainRNA(filename)
print ("BUILDSTRUCTURE: cg is ", cg)
sm=fbm.SpatialModel(cg)
print ("BUILDSTRUCTURE: sampling ")
sm.sample_stats()
print ("BUILDSTRUCTURE: traverse and build ")
sm.traverse_and_build()
print("BUILDSTRUCTURE: Writing File")
centroid0 = ftuv.get_vector_centroid(ftug.bg_virtual_residues(sm.bg))
for k in cg.coords.keys():
cg.coords[k] = ((cg.coords[k][0] - centroid0),
(cg.coords[k][1] - centroid0))
sm.bg.to_cg_file(filename)
print ("BUILDSTRUCTURE: DONE")
def main(args):
cgs = []
projs = []
for file_ in args.files:
cgs.append(ftmc.CoarseGrainRNA(file_))
try:
projs.append(ftmp.Projection2D(cgs[-1]))
except ValueError:
projs.append(None)
p_rmsds = OrderedDict()
cg_rmsds = OrderedDict()
p_rmsds[0] = 0.
cg_rmsds[0] = 0.
if "," in args.max_diff:
diffs = map(int, args.max_diff.split(","))
else:
diffs = range(1, int(args.max_diff) + 1)
for diff in diffs:
print("diff {}".format(diff))
prmsd = 0
cgrmsd = 0
count = 0
pcount = 0
for i in range(0, len(cgs) - diff):
count += 1
try:
vrs1 = np.array([
x for p in sorted(projs[i]._coords.keys())
for x in projs[i]._coords[p]
])
vrs2 = np.array([
x for p in sorted(projs[i + diff]._coords.keys())
for x in projs[i + diff]._coords[p]
])
prmsd += ftms.rmsd(vrs1, vrs2)
pcount += 1
except:
pass
cgrmsd += ftms.cg_rmsd(cgs[i], cgs[i + diff])
if count:
cg_rmsds[diff] = cgrmsd / count
if pcount:
p_rmsds[diff] = prmsd / pcount
print("projection RMSDs:", p_rmsds)
print("3D-RMSDs:", cg_rmsds)
import matplotlib.pyplot as plt
if args.target_structure and args.hausdorff_reference:
fig, (ax, axT, axH) = plt.subplots(3)
elif args.target_structure:
fig, (ax, axT) = plt.subplots(2)
elif args.hausdorff_reference:
fig, (ax, axH) = plt.subplots(2)
else:
fig, ax = plt.subplots()
if args.target_structure:
target = ftmc.CoarseGrainRNA(args.target_structure)
target_rmsds = []
target_proj_rmsds = []
try:
target_proj = ftmp.Projection2D(target)
except ValueError:
pass
else:
target_vrs = np.array([
x for p in sorted(target_proj._coords.keys())
for x in target_proj._coords[p]
])
for proj in projs:
try:
vrs1 = np.array([
x for p in sorted(proj._coords.keys())
for x in proj._coords[p]
])
prmsd = ftms.rmsd(vrs1, target_vrs)
target_proj_rmsds.append(prmsd)
except:
target_proj_rmsds.append(float("nan"))
target_vrs = ftug.bg_virtual_residues(target)
for cg in cgs:
vrs1 = ftug.bg_virtual_residues(cg)
cgrmsd = ftms.rmsd(vrs1, target_vrs)
target_rmsds.append(cgrmsd)
xval = np.arange(len(cgs)) * args.step_size
if target_proj_rmsds:
axT.plot(xval,
target_proj_rmsds,
label="Projection",
color="green")
axT.plot(xval, target_rmsds, label="3D", color="blue")
axT.set_title("RMSD to target structure")
axT.set_xlabel("step")
axT.set_ylabel("RMSD")
leg = axT.legend(framealpha=0.8, fancybox=True)
leg.get_frame().set_linewidth(0.0)
plt.subplots_adjust(hspace=0.4)
if args.hausdorff_reference:
ref_img = scipy.ndimage.imread(args.hausdorff_reference)
ref_quarter = (scipy.ndimage.zoom(ref_img, 0.3) > 150)
degrees = get_refimg_longest_axis(ref_img)
global_optima = []
local_optima = []
for cg in cgs:
score, img, params = fph.try_parameters(ref_img,
args.hausdorff_scale, cg,
degrees)
local_optima.append(score)
s, i, params = fph.globally_minimal_distance(
ref_quarter,
args.hausdorff_scale,
cg,
start_points=40,
starting_rotations=degrees,
virtual_atoms=False)
score, img, params = fph.locally_minimal_distance(
ref_img,
args.hausdorff_scale,
cg,
params[1],
params[2],
params[0],
maxiter=200)
global_optima.append(score)
axH.plot(xval, global_optima, label="Global Minimization", color="red")
axH.plot(xval,
local_optima,
label="Local Minimization",
color="lavender")
axH.set_title("Hausdorff Distance to reference image")
axH.set_xlabel("step")
axH.set_ylabel("Hausdorff Distance")
leg = axH.legend(framealpha=0.8, fancybox=True)
leg.get_frame().set_linewidth(0.0)
plt.subplots_adjust(hspace=0.4)
p_rmsds = np.array(list(p_rmsds.items()))
cg_rmsds = np.array(list(cg_rmsds.items()))
ax.plot(p_rmsds[:, 0] * args.step_size,
p_rmsds[:, 1],
label="Projection",
color="green",
marker="o")
ax.plot(cg_rmsds[:, 0] * args.step_size,
cg_rmsds[:, 1],
label="3D",
color="blue",
marker="o")
ax.set_title("Average RMSD between structures X steps apart.")
ax.set_xlabel("steps apart")
ax.set_ylabel("RMSD")
leg = ax.legend(framealpha=0.8, fancybox=True)
leg.get_frame().set_linewidth(0.0)
fig.patch.set_facecolor('white')
if args.save_fig:
with open(args.save_fig, "w") as f:
pickle.dump(fig, f)
plt.show()
def update_statistics(self, energy_function, sm, prev_energy, tracking_energies = None, tracked_energies=None):
'''
Add a newly sampled structure to the set of statistics.
:param energy_function: The energy_function used to evaluate the structure.
:param sm: The spatial model that was sampled.
:param prev_energy: The evaluated (accepted) energy of the current step
:tracking_energyis: The energy_functions which are calculated for statistics, but not used for sampling.
:tracked_energies: The energy values of the tracking_energies.
'''
self.counter += 1
if self.energy_orig is None:
self.energy_orig = 0.
try:
self.sm_orig.bg.add_all_virtual_residues()
self.energy_orig = energy_function.eval_energy(self.sm_orig)
except KeyError:
# most likely no native structure was provided
pass
energy = prev_energy
#energy = energy_function.eval_energy(sm, background=True)
if energy_function.uses_background():
energy_nobg = energy_function.eval_energy(sm, background=False)
else:
energy_nobg=energy
mcc = None
if self.centers_orig is not None:
r = 0.
if not self.no_rmsd:
centers_new = ftug.bg_virtual_residues(sm.bg)
r = cbr.centered_rmsd(self.centers_orig, centers_new)
#r = cbr.drmsd(self.centers_orig, centers_new)
cm = self.confusion_matrix_calculator.evaluate(sm.bg)
mcc = ftme.mcc(cm)
else:
# no original coordinates provided so we can't calculate rmsds
r = 0.
dist = None
dist2 = None
cg = sm.bg
dists = []
for (self.dist1, self.dist2) in self.dists:
node1 = cg.get_node_from_residue_num(self.dist1)
node2 = cg.get_node_from_residue_num(self.dist2)
pos1, len1 = cg.get_position_in_element(self.dist1)
pos2, len2 = cg.get_position_in_element(self.dist2)
#fud.pv('node1, node2, pos1, pos2')
vec1 = cg.coords[node1][1] - cg.coords[node1][0]
vec2 = cg.coords[node2][1] - cg.coords[node2][0]
#mid1 = (cg.coords[node1][0] + cg.coords[node1][1]) / 2
#mid2 = (cg.coords[node2][0] + cg.coords[node2][1]) / 2
mid1 = cg.coords[node1][0] + pos1 * (vec1 / len1)
mid2 = cg.coords[node2][0] + pos2 * (vec2 / len2)
#fud.pv('mid1, mid2')
dists += [ftuv.vec_distance(mid1, mid2)]
#self.energy_rmsd_structs += [(energy, r, sm.bg)]
self.energy_rmsd_structs += [(energy_nobg, r, copy.deepcopy(sm.bg))]
#self.energy_rmsd_structs += [(energy, r, sm.bg.copy())]
sorted_energies = sorted(self.energy_rmsd_structs, key=lambda x: x[0])
self.energy_rmsd_structs = sorted_energies[:self.save_n_best]
if r > self.highest_rmsd:
self.highest_rmsd = r
if r < self.lowest_rmsd:
self.lowest_rmsd = r
lowest_energy = sorted_energies[0][0]
lowest_rmsd = sorted_energies[0][1]
'''
if energy == lowest_energy:
for key in sm.angle_defs:
print >>sys.stderr, key, str(sm.angle_defs[key])
'''
if not self.silent:
if self.verbose:
'''
for energy_func in energy_function.energies:
print energy_func.__class__.__name__, energy_func.eval_energy(sm)
'''
_, rog=fbe.length_and_rog(sm.bg)
#output_str = u"native_energy [{:s} {:d}]: {:3d} {:5.03g} {:5.3f} ROG: {:5.3f} | min:
output_str = u"native_energy [%s %d]: %3d %5.03g %5.3f ROG: %5.3f | min: %5.2f (%5.2f) %5.2f | extreme_rmsds: %5.2f %5.2f (%.2f)" % ( sm.bg.name, sm.bg.seq_length, self.counter, energy, r , rog, lowest_energy, self.energy_orig, lowest_rmsd, self.lowest_rmsd, self.highest_rmsd, energy_nobg)
output_str += " |"
# assume that the energy function is a combined energy
if isinstance(self.energy_function, fbe.CombinedEnergy):
for e in self.energy_function.iterate_energies():
if isinstance(e,fbe.DistanceExponentialEnergy):
output_str += " [clamp {},{}: {:.1f}]".format(e.from_elem,
e.to_elem,
e.get_distance(sm))
if tracked_energies and tracking_energies:
output_str += " | [tracked Energies]"
for i,e in enumerate(tracking_energies):
sn=e.shortname()
if len(sn)>12:
sn=sn[:9]+"..."
output_str += " [{}]: ".format(sn)
output_str += "%5.03g" % (tracked_energies[i])
elif tracking_energies:
output_str += " | [tracked Energies]"
for e in tracking_energies:
sn=e.shortname()
if len(sn)>12:
sn=sn[:9]+"..."
output_str += " [{}]: ".format(sn)
output_str += "%5.03g" % (e.eval_energy(sm))
if dist:
output_str += " | dist %.2f" % (dist)
for dist2 in dists:
if dist2 is not None:
output_str += " | [dist2: %.2f]" % (dist2)
if mcc is not None:
output_str += " | [mcc: %.3f]" % (mcc)
output_str += " [time: %.1f]" % (time.time() - self.creation_time)
#Print to both STDOUT and the log file.
if self.output_file != sys.stdout:
print (output_str.strip())
if self.output_file != None:
print(output_str, file=self.output_file)
self.output_file.flush()
self.update_plots(energy, r)
'''
if self.counter % 1000 == 0:
import pdb; pdb.set_trace()
'''
if self.counter % 10 == 0:
if not self.silent:
self.save_top(self.save_n_best, counter=self.counter)
if self.step_save > 0 and self.counter % self.step_save == 0:
#If a projection match energy was used, save the optimal projection direction to the file.
if isinstance(self.energy_function, fbe.CombinedEnergy):
for e in self.energy_function.iterate_energies():
if hasattr(e, "accepted_projDir"):
sm.bg.project_from=e.accepted_projDir
sm.bg.to_cg_file(os.path.join(cbc.Configuration.sampling_output_dir, 'step%06d.coord' % (self.counter)))