def exercise_percentile_based_spread():
from libtbx.math_utils import percentile_based_spread
import random
import math
n_points = 123456
deltas = []
for i in range(n_points):
x = random.gauss(100, 10)
deltas.append(x)
for i in range(1000):
x = random.gauss(300, 30)
deltas.append(x)
pbs = percentile_based_spread(deltas)
rmsd = math.sqrt(sum([x**2 for x in deltas]) / n_points)
assert (pbs > 100) and (pbs < rmsd)
# Test small list processing
assert percentile_based_spread([1, 1]) > 0
def exercise_percentile_based_spread() :
from libtbx.math_utils import percentile_based_spread
import random
import math
n_points = 123456
deltas = []
for i in range(n_points) :
x = random.gauss(100, 10)
deltas.append(x)
for i in range(1000) :
x = random.gauss(300, 30)
deltas.append(x)
pbs = percentile_based_spread(deltas)
rmsd = math.sqrt(sum([ x**2 for x in deltas]) / n_points)
assert (pbs > 100) and (pbs < rmsd)
# Test small list processing
assert percentile_based_spread([1,1]) > 0
def compare_ligands_impl(ligand,
reference_ligands,
max_distance_between_centers_of_mass=8.0,
exclude_hydrogens=True,
implicit_matching=False,
verbose=False,
quiet=False,
raise_sorry_if_no_matching_atoms=True,
out=sys.stdout):
"""
Given a target ligand and a list of reference ligands, return the RMSD(s)
for any ligand determined to be approximately equivalent. (Usually there
will be just one of these, but this allows for alternate conformations.)
"""
from scitbx.array_family import flex
from scitbx.matrix import col
matching = []
atoms_1 = ligand.atoms()
sites_1 = atoms_1.extract_xyz()
xyz_mean_1 = sites_1.mean()
for ligand_2 in reference_ligands:
sites_2 = ligand_2.atoms().extract_xyz()
xyz_mean_2 = sites_2.mean()
dxyz = abs(col(xyz_mean_1) - col(xyz_mean_2))
if (dxyz < max_distance_between_centers_of_mass):
matching.append(ligand_2)
rmsds = []
pbss = []
for ligand_2 in matching:
atoms_2 = ligand_2.atoms()
isel_1 = flex.size_t()
isel_2 = flex.size_t()
for i_seq, atom_1 in enumerate(ligand.atoms()):
if (atom_1.element.strip() in ["H", "D"]) and (exclude_hydrogens):
continue
for j_seq, atom_2 in enumerate(ligand_2.atoms()):
if (atom_1.name == atom_2.name):
isel_1.append(i_seq)
isel_2.append(j_seq)
break
if (len(isel_1) == 0):
if (implicit_matching):
print(
" warning: no atom name matches found - will guess equivalence from sites",
file=out)
# XXX this is embarrassing... needs to be much smarter
for i_seq, atom_1 in enumerate(ligand.atoms()):
if (atom_1.element.strip() in ["H", "D"
]) and (exclude_hydrogens):
continue
j_seq_best = None
name_best = None
dxyz_best = sys.maxsize
for j_seq, atom_2 in enumerate(ligand_2.atoms()):
if (atom_1.element == atom_2.element):
dxyz = abs(col(atom_1.xyz) - col(atom_2.xyz))
if (dxyz < dxyz_best):
j_seq_best = j_seq
name_best = atom_2.name
dxyz_best = dxyz
if (j_seq_best is not None):
print(" '%s' : '%s' (distance = %.2f)" %
(atom_1.name, name_best, dxyz_best),
file=out)
isel_1.append(i_seq)
isel_2.append(j_seq_best)
if (len(isel_1) == 0):
if (raise_sorry_if_no_matching_atoms):
raise Sorry("No matching atoms found!")
else:
print(" WARNING: no matching atoms found!", file=out)
return None
sites_1 = sites_1.select(isel_1)
sites_2 = ligand_2.atoms().extract_xyz().select(isel_2)
rmsd = sites_1.rms_difference(sites_2)
pbs = percentile_based_spread((sites_2 - sites_1).norms())
if (not quiet):
print(" '%s' matches '%s': atoms=%d rmsd=%.3f" %
(ligand.id_str(), ligand_2.id_str(), sites_1.size(), rmsd),
file=out)
rmsds.append(rmsd)
pbss.append(pbs)
if (verbose) and (not quiet):
atoms = ligand.atoms()
dxyz = (sites_2 - sites_1).norms()
for i_seq, j_seq in zip(isel_1, isel_2):
print(" %s: dxyz=%.2f" %
(atoms_1[i_seq].id_str(), dxyz[i_seq]),
file=out)
return rmsds, pbss
def compare_ligands_impl (ligand,
reference_ligands,
max_distance_between_centers_of_mass=8.0,
exclude_hydrogens=True,
implicit_matching=False,
verbose=False,
quiet=False,
raise_sorry_if_no_matching_atoms=True,
out=sys.stdout) :
"""
Given a target ligand and a list of reference ligands, return the RMSD(s)
for any ligand determined to be approximately equivalent. (Usually there
will be just one of these, but this allows for alternate conformations.)
"""
from scitbx.array_family import flex
from scitbx.matrix import col
matching = []
atoms_1 = ligand.atoms()
sites_1 = atoms_1.extract_xyz()
xyz_mean_1 = sites_1.mean()
for ligand_2 in reference_ligands :
sites_2 = ligand_2.atoms().extract_xyz()
xyz_mean_2 = sites_2.mean()
dxyz = abs(col(xyz_mean_1) - col(xyz_mean_2))
if (dxyz < max_distance_between_centers_of_mass) :
matching.append(ligand_2)
rmsds = []
pbss = []
for ligand_2 in matching :
atoms_2 = ligand_2.atoms()
isel_1 = flex.size_t()
isel_2 = flex.size_t()
for i_seq, atom_1 in enumerate(ligand.atoms()) :
if (atom_1.element.strip() in ["H","D"]) and (exclude_hydrogens) :
continue
for j_seq, atom_2 in enumerate(ligand_2.atoms()) :
if (atom_1.name == atom_2.name) :
isel_1.append(i_seq)
isel_2.append(j_seq)
break
if (len(isel_1) == 0) :
if (implicit_matching) :
print >> out, " warning: no atom name matches found - will guess equivalence from sites"
# XXX this is embarrassing... needs to be much smarter
for i_seq, atom_1 in enumerate(ligand.atoms()) :
if (atom_1.element.strip() in ["H","D"]) and (exclude_hydrogens) :
continue
j_seq_best = None
name_best = None
dxyz_best = sys.maxint
for j_seq, atom_2 in enumerate(ligand_2.atoms()) :
if (atom_1.element == atom_2.element) :
dxyz = abs(col(atom_1.xyz) - col(atom_2.xyz))
if (dxyz < dxyz_best) :
j_seq_best = j_seq
name_best = atom_2.name
dxyz_best = dxyz
if (j_seq_best is not None) :
print >> out, " '%s' : '%s' (distance = %.2f)" % (atom_1.name,
name_best, dxyz_best)
isel_1.append(i_seq)
isel_2.append(j_seq_best)
if (len(isel_1) == 0) :
if (raise_sorry_if_no_matching_atoms) :
raise Sorry("No matching atoms found!")
else :
print >> out, " WARNING: no matching atoms found!"
return None
sites_1 = sites_1.select(isel_1)
sites_2 = ligand_2.atoms().extract_xyz().select(isel_2)
rmsd = sites_1.rms_difference(sites_2)
pbs = percentile_based_spread((sites_2 - sites_1).norms())
if (not quiet) :
print >> out, " '%s' matches '%s': atoms=%d rmsd=%.3f" % (
ligand.id_str(), ligand_2.id_str(), sites_1.size(), rmsd)
rmsds.append(rmsd)
pbss.append(pbs)
if (verbose) and (not quiet) :
atoms = ligand.atoms()
dxyz = (sites_2 - sites_1).norms()
for i_seq, j_seq in zip(isel_1, isel_2) :
print >> out, " %s: dxyz=%.2f" % (atoms_1[i_seq].id_str(),
dxyz[i_seq])
return rmsds, pbss