def exercise_isotropic_adp():
i_seqs = (0,)
weight = 2
u_cart = ((1,2,3,5,2,8),)
u_iso = (0,)
use_u_aniso = (True,)
p = adp_restraints.isotropic_adp_proxy(
i_seqs=i_seqs,
weight=weight)
assert p.i_seqs == i_seqs
assert approx_equal(p.weight, weight)
i = adp_restraints.isotropic_adp(u_cart=u_cart[0], weight=weight)
expected_deltas = (-1, 0, 1, 5, 2, 8)
expected_gradients = (-4, 0, 4, 40, 16, 64)
assert approx_equal(i.weight, weight)
assert approx_equal(i.deltas(), expected_deltas)
assert approx_equal(i.rms_deltas(), 4.5704364002673632)
assert approx_equal(i.residual(), 376.0)
assert approx_equal(i.gradients(), expected_gradients)
gradients_aniso_cart = flex.sym_mat3_double(1, (0,0,0,0,0,0))
gradients_iso = flex.double(1,0)
proxies = adp_restraints.shared_isotropic_adp_proxy([p,p])
u_cart = flex.sym_mat3_double(u_cart)
u_iso = flex.double(u_iso)
use_u_aniso = flex.bool(use_u_aniso)
params = adp_restraint_params(u_cart=u_cart, u_iso=u_iso, use_u_aniso=use_u_aniso)
residuals = adp_restraints.isotropic_adp_residuals(params, proxies=proxies)
assert approx_equal(residuals, (i.residual(),i.residual()))
deltas_rms = adp_restraints.isotropic_adp_deltas_rms(params, proxies=proxies)
assert approx_equal(deltas_rms, (i.rms_deltas(),i.rms_deltas()))
residual_sum = adp_restraints.isotropic_adp_residual_sum(
params,
proxies=proxies,
gradients_aniso_cart=gradients_aniso_cart
)
assert approx_equal(residual_sum, 752.0)
fd_grads_aniso, fd_grads_iso = finite_difference_gradients(
restraint_type=adp_restraints.isotropic_adp,
proxy=p,
u_cart=u_cart,
u_iso=u_iso,
use_u_aniso=use_u_aniso
)
for g,e in zip(gradients_aniso_cart, fd_grads_aniso):
assert approx_equal(g, matrix.col(e)*2)
#
# check frame invariance of residual
#
u_cart = matrix.sym(sym_mat3=(0.1,0.2,0.05,0.03,0.02,0.01))
a = adp_restraints.isotropic_adp(
u_cart=u_cart.as_sym_mat3(), weight=1)
expected_residual = a.residual()
gen = flex.mersenne_twister()
for i in range(20):
R = matrix.rec(gen.random_double_r3_rotation_matrix(),(3,3))
u_cart_rot = R * u_cart * R.transpose()
a = adp_restraints.isotropic_adp(
u_cart=u_cart_rot.as_sym_mat3(), weight=1)
assert approx_equal(a.residual(), expected_residual)
def exercise_isotropic_adp():
i_seqs = (0,)
weight = 2
u_cart = ((1,2,3,5,2,8),)
u_iso = (0,)
use_u_aniso = (True,)
p = adp_restraints.isotropic_adp_proxy(
i_seqs=i_seqs,
weight=weight)
assert p.i_seqs == i_seqs
assert approx_equal(p.weight, weight)
i = adp_restraints.isotropic_adp(u_cart=u_cart[0], weight=weight)
expected_deltas = (-1, 0, 1, 5, 2, 8)
expected_gradients = (-4, 0, 4, 40, 16, 64)
assert approx_equal(i.weight, weight)
assert approx_equal(i.deltas(), expected_deltas)
assert approx_equal(i.rms_deltas(), 4.5704364002673632)
assert approx_equal(i.residual(), 376.0)
assert approx_equal(i.gradients(), expected_gradients)
gradients_aniso_cart = flex.sym_mat3_double(1, (0,0,0,0,0,0))
gradients_iso = flex.double(1,0)
proxies = adp_restraints.shared_isotropic_adp_proxy([p,p])
u_cart = flex.sym_mat3_double(u_cart)
u_iso = flex.double(u_iso)
use_u_aniso = flex.bool(use_u_aniso)
params = adp_restraint_params(u_cart=u_cart, u_iso=u_iso, use_u_aniso=use_u_aniso)
residuals = adp_restraints.isotropic_adp_residuals(params, proxies=proxies)
assert approx_equal(residuals, (i.residual(),i.residual()))
deltas_rms = adp_restraints.isotropic_adp_deltas_rms(params, proxies=proxies)
assert approx_equal(deltas_rms, (i.rms_deltas(),i.rms_deltas()))
residual_sum = adp_restraints.isotropic_adp_residual_sum(
params,
proxies=proxies,
gradients_aniso_cart=gradients_aniso_cart
)
assert approx_equal(residual_sum, 752.0)
fd_grads_aniso, fd_grads_iso = finite_difference_gradients(
restraint_type=adp_restraints.isotropic_adp,
proxy=p,
u_cart=u_cart,
u_iso=u_iso,
use_u_aniso=use_u_aniso
)
for g,e in zip(gradients_aniso_cart, fd_grads_aniso):
assert approx_equal(g, matrix.col(e)*2)
#
# check frame invariance of residual
#
u_cart = matrix.sym(sym_mat3=(0.1,0.2,0.05,0.03,0.02,0.01))
a = adp_restraints.isotropic_adp(
u_cart=u_cart.as_sym_mat3(), weight=1)
expected_residual = a.residual()
gen = flex.mersenne_twister()
for i in range(20):
R = matrix.rec(gen.random_double_r3_rotation_matrix(),(3,3))
u_cart_rot = R * u_cart * R.transpose()
a = adp_restraints.isotropic_adp(
u_cart=u_cart_rot.as_sym_mat3(), weight=1)
assert approx_equal(a.residual(), expected_residual)
def __init__(self,
u_cart,
u_iso=None,
use_u_aniso=None,
sites_cart=None,
adp_similarity_proxies=None,
rigid_bond_proxies=None,
isotropic_adp_proxies=None,
compute_gradients=True,
gradients_aniso_cart=None,
gradients_iso=None,
disable_asu_cache=False,
normalization=False):
adopt_init_args(self, locals())
self.number_of_restraints = 0
self.residual_sum = 0
self.normalization_factor = None
if (adp_similarity_proxies is not None):
assert u_iso is not None and use_u_aniso is not None
if (rigid_bond_proxies is not None): assert sites_cart is not None
if (sites_cart is not None): assert sites_cart.size() == u_cart.size()
if (u_iso is not None): assert u_iso.size() == u_cart.size()
if (use_u_aniso is not None):
assert use_u_aniso.size() == u_cart.size()
if (compute_gradients):
if (self.gradients_aniso_cart is None):
self.gradients_aniso_cart = flex.sym_mat3_double(
sites_cart.size(), [0, 0, 0, 0, 0, 0])
else:
assert self.gradients_aniso_cart.size() == sites_cart.size()
if (u_iso is not None and self.gradients_iso is None):
self.gradients_iso = flex.double(sites_cart.size(), 0)
elif (u_iso is not None):
assert self.gradients_iso.size() == sites_cart.size()
if (adp_similarity_proxies is None):
self.n_adp_similarity_proxies = None
self.adp_similarity_residual_sum = 0
else:
self.n_adp_similarity_proxies = len(adp_similarity_proxies)
self.adp_similarity_residual_sum = adp_restraints.adp_similarity_residual_sum(
u_cart=u_cart,
u_iso=u_iso,
use_u_aniso=use_u_aniso,
proxies=adp_similarity_proxies,
gradients_aniso_cart=self.gradients_aniso_cart,
gradients_iso=self.gradients_iso)
self.number_of_restraints += 6 * self.n_adp_similarity_proxies
self.residual_sum += self.adp_similarity_residual_sum
if (rigid_bond_proxies is None):
self.n_rigid_bond_proxies = None
self.rigid_bond_residual_sum = 0
else:
self.n_rigid_bond_proxies = len(rigid_bond_proxies)
self.rigid_bond_residual_sum = adp_restraints.rigid_bond_residual_sum(
sites_cart=sites_cart,
u_cart=u_cart,
proxies=rigid_bond_proxies,
gradients_aniso_cart=self.gradients_aniso_cart)
self.number_of_restraints += self.n_rigid_bond_proxies
self.residual_sum += self.rigid_bond_residual_sum
if (isotropic_adp_proxies is None):
self.n_isotropic_adp_proxies = None
self.isotropic_adp_residual_sum = 0
else:
self.n_isotropic_adp_proxies = len(isotropic_adp_proxies)
self.isotropic_adp_residual_sum = adp_restraints.isotropic_adp_residual_sum(
u_cart=u_cart,
proxies=isotropic_adp_proxies,
gradients_aniso_cart=self.gradients_aniso_cart)
self.number_of_restraints += self.n_isotropic_adp_proxies
self.residual_sum += self.isotropic_adp_residual_sum
self.finalize_target_and_gradients()
def __init__(self, u_cart,
u_iso=None,
use_u_aniso=None,
sites_cart=None,
adp_similarity_proxies=None,
rigid_bond_proxies=None,
isotropic_adp_proxies=None,
compute_gradients=True,
gradients_aniso_cart=None,
gradients_iso=None,
disable_asu_cache=False,
normalization=False):
adopt_init_args(self, locals())
self.number_of_restraints = 0
self.residual_sum = 0
self.normalization_factor = None
if (adp_similarity_proxies is not None):
assert u_iso is not None and use_u_aniso is not None
if (rigid_bond_proxies is not None): assert sites_cart is not None
if (sites_cart is not None): assert sites_cart.size() == u_cart.size()
if (u_iso is not None): assert u_iso.size() == u_cart.size()
if (use_u_aniso is not None): assert use_u_aniso.size() == u_cart.size()
if (compute_gradients):
if (self.gradients_aniso_cart is None):
self.gradients_aniso_cart = flex.sym_mat3_double(
sites_cart.size(), [0,0,0,0,0,0])
else:
assert self.gradients_aniso_cart.size() == sites_cart.size()
if (u_iso is not None and self.gradients_iso is None):
self.gradients_iso = flex.double(sites_cart.size(), 0)
elif (u_iso is not None):
assert self.gradients_iso.size() == sites_cart.size()
if (adp_similarity_proxies is None):
self.n_adp_similarity_proxies = None
self.adp_similarity_residual_sum = 0
else:
self.n_adp_similarity_proxies = len(adp_similarity_proxies)
self.adp_similarity_residual_sum = adp_restraints.adp_similarity_residual_sum(
u_cart=u_cart,
u_iso=u_iso,
use_u_aniso=use_u_aniso,
proxies=adp_similarity_proxies,
gradients_aniso_cart=self.gradients_aniso_cart,
gradients_iso=self.gradients_iso)
self.number_of_restraints += 6 * self.n_adp_similarity_proxies
self.residual_sum += self.adp_similarity_residual_sum
if (rigid_bond_proxies is None):
self.n_rigid_bond_proxies = None
self.rigid_bond_residual_sum = 0
else:
self.n_rigid_bond_proxies = len(rigid_bond_proxies)
self.rigid_bond_residual_sum = adp_restraints.rigid_bond_residual_sum(
sites_cart=sites_cart,
u_cart=u_cart,
proxies=rigid_bond_proxies,
gradients_aniso_cart=self.gradients_aniso_cart)
self.number_of_restraints += self.n_rigid_bond_proxies
self.residual_sum += self.rigid_bond_residual_sum
if (isotropic_adp_proxies is None):
self.n_isotropic_adp_proxies = None
self.isotropic_adp_residual_sum = 0
else:
self.n_isotropic_adp_proxies = len(isotropic_adp_proxies)
self.isotropic_adp_residual_sum = adp_restraints.isotropic_adp_residual_sum(
u_cart=u_cart,
proxies=isotropic_adp_proxies,
gradients_aniso_cart=self.gradients_aniso_cart)
self.number_of_restraints += self.n_isotropic_adp_proxies
self.residual_sum += self.isotropic_adp_residual_sum
self.finalize_target_and_gradients()
