def adp_similarity_deviation(self):
if (self.n_adp_similarity_proxies is not None):
adp_similarity_deltas_rms = adp_restraints.adp_similarity_deltas_rms(
u_cart=self.u_cart,
u_iso=self.u_iso,
use_u_aniso=self.use_u_aniso,
proxies=self.adp_similarity_proxies)
a_sq = adp_similarity_deltas_rms * adp_similarity_deltas_rms
a_ave = math.sqrt(flex.mean_default(a_sq, 0))
a_max = math.sqrt(flex.max_default(a_sq, 0))
a_min = math.sqrt(flex.min_default(a_sq, 0))
return a_min, a_max, a_ave
def adp_similarity_deviation(self):
if (self.n_adp_similarity_proxies is not None):
adp_similarity_deltas_rms = adp_restraints.adp_similarity_deltas_rms(
u_cart=self.u_cart,
u_iso=self.u_iso,
use_u_aniso=self.use_u_aniso,
proxies=self.adp_similarity_proxies)
a_sq = adp_similarity_deltas_rms * adp_similarity_deltas_rms
a_ave = math.sqrt(flex.mean_default(a_sq, 0))
a_max = math.sqrt(flex.max_default(a_sq, 0))
a_min = math.sqrt(flex.min_default(a_sq, 0))
return a_min, a_max, a_ave
def exercise_adp_similarity():
u_cart = ((1,3,2,4,3,6),(2,4,2,6,5,1))
u_iso = (-1,-1)
use_u_aniso = (True, True)
weight = 1
a = adp_restraints.adp_similarity(
u_cart=u_cart,
weight=weight)
assert approx_equal(a.use_u_aniso, use_u_aniso)
assert a.weight == weight
assert approx_equal(a.residual(), 68)
assert approx_equal(a.gradients2(),
((-2.0, -2.0, 0.0, -8.0, -8.0, 20.0), (2.0, 2.0, -0.0, 8.0, 8.0, -20.0)))
assert approx_equal(a.deltas(), (-1.0, -1.0, 0.0, -2.0, -2.0, 5.0))
assert approx_equal(a.rms_deltas(), 2.7487370837451071)
#
u_cart = ((1,3,2,4,3,6),(-1,-1,-1,-1,-1,-1))
u_iso = (-1,2)
use_u_aniso = (True, False)
a = adp_restraints.adp_similarity(
u_cart[0], u_iso[1], weight=weight)
assert approx_equal(a.use_u_aniso, use_u_aniso)
assert a.weight == weight
assert approx_equal(a.residual(), 124)
assert approx_equal(a.gradients2(),
((-2, 2, 0, 16, 12, 24), (2, -2, 0, -16, -12, -24)))
assert approx_equal(a.deltas(), (-1, 1, 0, 4, 3, 6))
assert approx_equal(a.rms_deltas(), 3.711842908553348)
#
i_seqs_aa = (1,2) # () - ()
i_seqs_ai = (1,0) # () - o
i_seqs_ia = (3,2) # o - ()
i_seqs_ii = (0,3) # o - o
p_aa = adp_restraints.adp_similarity_proxy(i_seqs=i_seqs_aa,weight=weight)
p_ai = adp_restraints.adp_similarity_proxy(i_seqs=i_seqs_ai,weight=weight)
p_ia = adp_restraints.adp_similarity_proxy(i_seqs=i_seqs_ia,weight=weight)
p_ii = adp_restraints.adp_similarity_proxy(i_seqs=i_seqs_ii,weight=weight)
assert p_aa.i_seqs == i_seqs_aa
assert p_aa.weight == weight
u_cart = flex.sym_mat3_double(((-1,-1,-1,-1,-1,-1),
(1,2,2,4,3,6),
(2,4,2,6,5,1),
(-1,-1,-1,-1,-1,-1)))
u_iso = flex.double((1,-1,-1,2))
use_u_aniso = flex.bool((False, True,True,False))
for p in (p_aa,p_ai,p_ia,p_ii):
params = adp_restraint_params(u_cart=u_cart, u_iso=u_iso, use_u_aniso=use_u_aniso)
a = adp_restraints.adp_similarity(params, proxy=p)
assert approx_equal(a.weight, weight)
#
gradients_aniso_cart = flex.sym_mat3_double(u_cart.size(), (0,0,0,0,0,0))
gradients_iso = flex.double(u_cart.size(), 0)
proxies = adp_restraints.shared_adp_similarity_proxy([p,p])
residuals = adp_restraints.adp_similarity_residuals(params, proxies=proxies)
assert approx_equal(residuals, (a.residual(),a.residual()))
deltas_rms = adp_restraints.adp_similarity_deltas_rms(params, proxies=proxies)
assert approx_equal(deltas_rms, (a.rms_deltas(),a.rms_deltas()))
residual_sum = adp_restraints.adp_similarity_residual_sum(
params,
proxies=proxies,
gradients_aniso_cart=gradients_aniso_cart,
gradients_iso=gradients_iso)
assert approx_equal(residual_sum, 2 * a.residual())
fd_grads_aniso, fd_grads_iso = finite_difference_gradients(
restraint_type=adp_restraints.adp_similarity,
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)
for g,e in zip(gradients_iso, fd_grads_iso):
assert approx_equal(g, e*2)
#
# check frame invariance of residual
#
u_cart_1 = matrix.sym(sym_mat3=(0.1,0.2,0.05,0.03,0.02,0.01))
u_cart_2 = matrix.sym(sym_mat3=(0.21,0.32,0.11,0.02,0.02,0.07))
u_cart = (u_cart_1.as_sym_mat3(),u_cart_2.as_sym_mat3())
u_iso = (-1, -1)
use_u_aniso = (True, True)
a = adp_restraints.adp_similarity(u_cart, 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_1_rot = R * u_cart_1 * R.transpose()
u_cart_2_rot = R * u_cart_2 * R.transpose()
u_cart = (u_cart_1_rot.as_sym_mat3(),u_cart_2_rot.as_sym_mat3())
a = adp_restraints.adp_similarity(u_cart, weight=1)
assert approx_equal(a.residual(), expected_residual)
def exercise_adp_similarity():
u_cart = ((1,3,2,4,3,6),(2,4,2,6,5,1))
u_iso = (-1,-1)
use_u_aniso = (True, True)
weight = 1
a = adp_restraints.adp_similarity(
u_cart=u_cart,
weight=weight)
assert approx_equal(a.use_u_aniso, use_u_aniso)
assert a.weight == weight
assert approx_equal(a.residual(), 68)
assert approx_equal(a.gradients2(),
((-2.0, -2.0, 0.0, -8.0, -8.0, 20.0), (2.0, 2.0, -0.0, 8.0, 8.0, -20.0)))
assert approx_equal(a.deltas(), (-1.0, -1.0, 0.0, -2.0, -2.0, 5.0))
assert approx_equal(a.rms_deltas(), 2.7487370837451071)
#
u_cart = ((1,3,2,4,3,6),(-1,-1,-1,-1,-1,-1))
u_iso = (-1,2)
use_u_aniso = (True, False)
a = adp_restraints.adp_similarity(
u_cart[0], u_iso[1], weight=weight)
assert approx_equal(a.use_u_aniso, use_u_aniso)
assert a.weight == weight
assert approx_equal(a.residual(), 124)
assert approx_equal(a.gradients2(),
((-2, 2, 0, 16, 12, 24), (2, -2, 0, -16, -12, -24)))
assert approx_equal(a.deltas(), (-1, 1, 0, 4, 3, 6))
assert approx_equal(a.rms_deltas(), 3.711842908553348)
#
i_seqs_aa = (1,2) # () - ()
i_seqs_ai = (1,0) # () - o
i_seqs_ia = (3,2) # o - ()
i_seqs_ii = (0,3) # o - o
p_aa = adp_restraints.adp_similarity_proxy(i_seqs=i_seqs_aa,weight=weight)
p_ai = adp_restraints.adp_similarity_proxy(i_seqs=i_seqs_ai,weight=weight)
p_ia = adp_restraints.adp_similarity_proxy(i_seqs=i_seqs_ia,weight=weight)
p_ii = adp_restraints.adp_similarity_proxy(i_seqs=i_seqs_ii,weight=weight)
assert p_aa.i_seqs == i_seqs_aa
assert p_aa.weight == weight
u_cart = flex.sym_mat3_double(((-1,-1,-1,-1,-1,-1),
(1,2,2,4,3,6),
(2,4,2,6,5,1),
(-1,-1,-1,-1,-1,-1)))
u_iso = flex.double((1,-1,-1,2))
use_u_aniso = flex.bool((False, True,True,False))
for p in (p_aa,p_ai,p_ia,p_ii):
params = adp_restraint_params(u_cart=u_cart, u_iso=u_iso, use_u_aniso=use_u_aniso)
a = adp_restraints.adp_similarity(params, proxy=p)
assert approx_equal(a.weight, weight)
#
gradients_aniso_cart = flex.sym_mat3_double(u_cart.size(), (0,0,0,0,0,0))
gradients_iso = flex.double(u_cart.size(), 0)
proxies = adp_restraints.shared_adp_similarity_proxy([p,p])
residuals = adp_restraints.adp_similarity_residuals(params, proxies=proxies)
assert approx_equal(residuals, (a.residual(),a.residual()))
deltas_rms = adp_restraints.adp_similarity_deltas_rms(params, proxies=proxies)
assert approx_equal(deltas_rms, (a.rms_deltas(),a.rms_deltas()))
residual_sum = adp_restraints.adp_similarity_residual_sum(
params,
proxies=proxies,
gradients_aniso_cart=gradients_aniso_cart,
gradients_iso=gradients_iso)
assert approx_equal(residual_sum, 2 * a.residual())
fd_grads_aniso, fd_grads_iso = finite_difference_gradients(
restraint_type=adp_restraints.adp_similarity,
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)
for g,e in zip(gradients_iso, fd_grads_iso):
assert approx_equal(g, e*2)
#
# check frame invariance of residual
#
u_cart_1 = matrix.sym(sym_mat3=(0.1,0.2,0.05,0.03,0.02,0.01))
u_cart_2 = matrix.sym(sym_mat3=(0.21,0.32,0.11,0.02,0.02,0.07))
u_cart = (u_cart_1.as_sym_mat3(),u_cart_2.as_sym_mat3())
u_iso = (-1, -1)
use_u_aniso = (True, True)
a = adp_restraints.adp_similarity(u_cart, 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_1_rot = R * u_cart_1 * R.transpose()
u_cart_2_rot = R * u_cart_2 * R.transpose()
u_cart = (u_cart_1_rot.as_sym_mat3(),u_cart_2_rot.as_sym_mat3())
a = adp_restraints.adp_similarity(u_cart, weight=1)
assert approx_equal(a.residual(), expected_residual)