def test_MMVT_data_sample_fill_out_data_quantities():
model = make_simple_model()
N_alpha_beta = {(0,1):12, (1,0):12,
(1,2):12, (2,1):12,
(2,3):6, (3,2):6}
k_alpha_beta = {(0,1):20.0, (1,0):10.0,
(1,2):10.0, (2,1):(40.0/3.0),
(2,3):(20.0/3.0), (3,2):20.0}
N_i_j_alpha = [{},
{(0,1):4, (1,0):4},
{(1,2):2, (2,1):2},
{}]
R_i_alpha_total = [{0: 1.2},
{0: 1.2, 1:1.2},
{1: 1.2, 2:0.6},
{2: 0.6}]
T_alpha_total = [1.2,
2.4,
1.8,
0.6]
main_data_sample = mmvt_analyze.MMVT_data_sample(
model, N_alpha_beta, k_alpha_beta, N_i_j_alpha,
R_i_alpha_total, T_alpha_total)
main_data_sample.calculate_pi_alpha()
main_data_sample.fill_out_data_quantities()
N_ij = {(0,1): 1, (1,0): 1, (1,2): 0.5, (2,1): 0.5}
R_i = {0: 0.6, 1: 0.6, 2: 0.3}
compare_dicts(N_ij, main_data_sample.N_ij)
compare_dicts(R_i, main_data_sample.R_i)
return
def test_MMVT_data_sample_calculate_pi_alpha():
model = make_simple_model()
N_alpha_beta = {(0,1):10, (1,0):10,
(1,2):10, (2,1):10,
(2,3):10, (3,2):10}
k_alpha_beta = {(0,1):4.0, (1,0):4.0,
(1,2):4.0, (2,1):4.0,
(2,3):4.0, (3,2):4.0}
N_i_j_alpha = [{},
{(0,1):3, (1,0):3},
{(1,2):3, (2,1):3},
{}]
R_i_alpha_total = [{0: 8.0},
{0: 8.0, 1:8.0},
{1: 8.0, 2:8.0},
{2: 8.0}]
T_alpha_total = [2.5,
2.5,
2.5,
2.5]
main_data_sample = mmvt_analyze.MMVT_data_sample(
model, N_alpha_beta, k_alpha_beta, N_i_j_alpha,
R_i_alpha_total, T_alpha_total)
main_data_sample.calculate_pi_alpha()
assert np.all(np.isclose(main_data_sample.pi_alpha[:-1],
main_data_sample.pi_alpha[0]))
assert np.isclose(main_data_sample.pi_alpha[-1], 0.0)
return
def test_MMVT_data_sample_fill_k_from_matrices():
model = make_simple_model()
k_alpha_beta_exp = {(0,1):20.0, (1,0):10.0,
(1,2):10.0, (2,1):(40.0/3.0),
(2,3):(20.0/3.0), (3,2):20.0}
k_alpha_beta_matrix = np.array([
[-20.0, 20.0, 0.0, 0.0],
[10.0, -20.0, 10.0, 0.0],
[0.0, (40.0/3.0), -(40.0/3.0)-(20.0/3.0), (20.0/3.0)],
[0.0, 0.0, 20.0, -20.0]])
main_data_sample = mmvt_analyze.MMVT_data_sample(
model, None, None, None,
None, None)
main_data_sample.fill_k_from_matrices(k_alpha_beta_matrix)
compare_dicts(k_alpha_beta_exp, main_data_sample.k_alpha_beta)
return
def fill_out_data_samples_mmvt(self):
"""
Now that the statistics for each anchor have been extracted
from the output files, construct the global transition
statistics objects. Applies to systems using MMVT milestoning.
"""
N_alpha_beta = defaultdict(int)
k_alpha_beta = defaultdict(float)
N_i_j_alpha = []
R_i_alpha_total = []
T_alpha_total = []
for alpha, anchor1 in enumerate(self.model.anchors):
if anchor1.bulkstate:
continue
anchor_N_alpha_beta = self.anchor_stats_list[alpha].N_alpha_beta
anchor_k_alpha_beta = self.anchor_stats_list[alpha].k_alpha_beta
for beta, anchor2 in enumerate(self.model.anchors):
if anchor2.bulkstate:
continue
if alpha == beta:
continue
id_alias = anchor1.alias_from_neighbor_id(anchor2.index)
if id_alias is None:
continue
if len(anchor_N_alpha_beta) == 0:
# Then no transitions were observed in this anchor
default_flux = FAST_FLUX
else:
default_flux = LOW_FLUX
if id_alias in anchor_N_alpha_beta:
N_alpha_beta[(alpha, beta)] = anchor_N_alpha_beta[id_alias]
k_alpha_beta[(alpha, beta)] = anchor_k_alpha_beta[id_alias]
else:
N_alpha_beta[(alpha, beta)] = 0
k_alpha_beta[(alpha, beta)] = default_flux #0.0
anchor_N_i_j_alpha = self.anchor_stats_list[alpha].N_i_j_alpha
N_i_j_alpha_element = defaultdict(int)
R_i_alpha_element = defaultdict(float)
# Fill out empty milestone statistics
for milestone1 in anchor1.milestones:
R_i_alpha_element[milestone1.index] = 0.0
for milestone2 in anchor1.milestones:
if milestone1.index == milestone2.index:
continue
new_key = (milestone1.index, milestone2.index)
N_i_j_alpha_element[new_key] = 0
for key in anchor_N_i_j_alpha:
(alias_id_i, alias_id_j) = key
id_i = anchor1.id_from_alias(alias_id_i)
id_j = anchor1.id_from_alias(alias_id_j)
new_key = (id_i, id_j)
N_i_j_alpha_element[new_key] = anchor_N_i_j_alpha[key]
N_i_j_alpha.append(N_i_j_alpha_element)
anchor_R_i_alpha = self.anchor_stats_list[alpha].R_i_alpha_total
for key in anchor_R_i_alpha:
alias_id_i = key
id_i = anchor1.id_from_alias(alias_id_i)
assert id_i is not None, \
"Anchor {} missing alias {}.".format(anchor1.index,
alias_id_i)
R_i_alpha_element[id_i] = anchor_R_i_alpha[key]
R_i_alpha_total.append(R_i_alpha_element)
anchor_T_alpha = self.anchor_stats_list[alpha].T_alpha_total
T_alpha_total.append(anchor_T_alpha)
self.main_data_sample = mmvt_analyze.MMVT_data_sample(
self.model, N_alpha_beta, k_alpha_beta, N_i_j_alpha,
R_i_alpha_total, T_alpha_total)
return
def fill_out_data_samples_mmvt(self):
"""
Now that the statistics for each anchor have been extracted
from the output files, construct the global transition
statistics objects. Applies to systems using MMVT milestoning.
"""
N_alpha_beta = defaultdict(int)
k_alpha_beta = defaultdict(float)
N_i_j_alpha = []
R_i_alpha_total = []
R_i_alpha_average = []
R_i_alpha_std_dev = []
R_i_alpha_count = []
T_alpha_total = []
T_alpha_average = []
T_alpha_std_dev = []
T_alpha_count = []
for alpha, anchor1 in enumerate(self.model.anchors):
if anchor1.bulkstate:
continue
anchor_N_alpha_beta = self.anchor_stats_list[alpha].N_alpha_beta
anchor_k_alpha_beta = self.anchor_stats_list[alpha].k_alpha_beta
for beta, anchor2 in enumerate(self.model.anchors):
if anchor2.bulkstate:
continue
if alpha == beta:
continue
id_alias = anchor1.alias_from_neighbor_id(anchor2.index)
if id_alias is None:
continue
if id_alias in anchor_N_alpha_beta:
N_alpha_beta[(alpha, beta)] = anchor_N_alpha_beta[id_alias]
k_alpha_beta[(alpha, beta)] = anchor_k_alpha_beta[id_alias]
else:
N_alpha_beta[(alpha, beta)] = 0
k_alpha_beta[(alpha, beta)] = 0.0
anchor_N_i_j_alpha = self.anchor_stats_list[alpha].N_i_j_alpha
N_i_j_alpha_element = defaultdict(int)
for key in anchor_N_i_j_alpha:
(alias_id_i, alias_id_j) = key
id_i = anchor1.id_from_alias(alias_id_i)
id_j = anchor1.id_from_alias(alias_id_j)
new_key = (id_i, id_j)
N_i_j_alpha_element[new_key] = anchor_N_i_j_alpha[key]
N_i_j_alpha.append(N_i_j_alpha_element)
anchor_R_i_alpha = self.anchor_stats_list[alpha].R_i_alpha_total
anchor_R_i_alpha_std = self.anchor_stats_list[
alpha].R_i_alpha_std_dev
anchor_R_i_alpha_list = self.anchor_stats_list[
alpha].R_i_alpha_list
R_i_alpha_element = defaultdict(float)
R_i_alpha_count_element = defaultdict(int)
R_i_alpha_std_element = defaultdict(float)
for key in anchor_R_i_alpha:
alias_id_i = key
id_i = anchor1.id_from_alias(alias_id_i)
R_i_alpha_element[id_i] = anchor_R_i_alpha[key]
R_i_alpha_std_element[id_i] = anchor_R_i_alpha_std[key]
R_i_alpha_count_element[id_i] = len(anchor_R_i_alpha_list[key])
R_i_alpha_total.append(R_i_alpha_element)
R_i_alpha_std_dev.append(R_i_alpha_std_element)
R_i_alpha_count.append(R_i_alpha_count_element)
anchor_T_alpha = self.anchor_stats_list[alpha].T_alpha_total
anchor_T_alpha_std = self.anchor_stats_list[alpha].T_alpha_std_dev
anchor_T_alpha_list = self.anchor_stats_list[alpha].T_alpha_list
T_alpha_total.append(anchor_T_alpha)
T_alpha_std_dev.append(anchor_T_alpha_std)
T_alpha_count.append(len(anchor_T_alpha_list))
self.main_data_sample = mmvt_analyze.MMVT_data_sample(
self.model, N_alpha_beta, k_alpha_beta, N_i_j_alpha,
R_i_alpha_total, T_alpha_total)
for i in range(self.num_error_samples):
sampled_k_alpha_beta, sampled_N_i_j_alpha, \
sampled_R_i_alpha_total, sampled_T_alpha_total \
= self.resample_k_N_R_T(
N_alpha_beta, N_i_j_alpha, R_i_alpha_total,
R_i_alpha_average, R_i_alpha_std_dev, R_i_alpha_count,
T_alpha_total, T_alpha_average, T_alpha_std_dev,
T_alpha_count)
data_sample = mmvt_analyze.MMVT_data_sample(
self.model, N_alpha_beta, sampled_k_alpha_beta,
sampled_N_i_j_alpha, sampled_R_i_alpha_total,
sampled_T_alpha_total)
self.data_sample_list.append(data_sample)
return
def test_mcmc_3x3_mmvt(tmpdir_factory):
"""
Use the same statistics to generate both MMVT and Elber rate matrices.
Then use the data to sample matrices and compare the distributions to
ensure that the MMVT and Elber matrix sampling methods are consistent.
"""
num = 10000 #100000
stride = 4
skip = 90
n_anchors = 4
n_milestones = 3
# generate data to feed directly into MMVT_data_sample()
model = base.Model()
model.num_milestones = n_milestones
model.num_anchors = n_anchors
anchor0 = mmvt_base.MMVT_toy_anchor()
anchor0.index = 0
milestone_0_0 = base.Milestone()
milestone_0_0.index = 0
milestone_0_0.neighbor_anchor_index = 1
milestone_0_0.alias_index = 1
anchor0.milestones = [milestone_0_0]
anchor1 = mmvt_base.MMVT_toy_anchor()
anchor1.index = 1
milestone_1_0 = base.Milestone()
milestone_1_0.index = 0
milestone_1_0.neighbor_anchor_index = 0
milestone_1_0.alias_index = 1
milestone_1_1 = base.Milestone()
milestone_1_1.index = 1
milestone_1_1.neighbor_anchor_index = 2
milestone_1_1.alias_index = 2
anchor1.milestones = [milestone_1_0, milestone_1_1]
anchor2 = mmvt_base.MMVT_toy_anchor()
anchor2.index = 2
milestone_2_0 = base.Milestone()
milestone_2_0.index = 1
milestone_2_0.neighbor_anchor_index = 1
milestone_2_0.alias_index = 1
milestone_2_1 = base.Milestone()
milestone_2_1.index = 2
milestone_2_1.neighbor_anchor_index = 3
milestone_2_1.alias_index = 2
anchor2.milestones = [milestone_2_0, milestone_2_1]
anchor3 = mmvt_base.MMVT_toy_anchor()
anchor3.index = 3
milestone_3_0 = base.Milestone()
milestone_3_0.index = 2
milestone_3_0.neighbor_anchor_index = 2
milestone_3_0.alias_index = 1
anchor3.milestones = [milestone_3_0]
model.anchors = [anchor0, anchor1, anchor2, anchor3]
# MMVT stats
N_alpha_beta = {(0,1):12, (1,0):12,
(1,2):12, (2,1):12,
(2,3):6, (3,2):6}
k_alpha_beta = {(0,1):20.0, (1,0):10.0,
(1,2):10.0, (2,1):(40.0/3.0),
(2,3):(20.0/3.0), (3,2):20.0}
N_i_j_alpha = [{},
{(0,1):4, (1,0):4},
{(1,2):2, (2,1):2},
{}]
R_i_alpha_total = [{0: 1.2},
{0: 1.2, 1:1.2},
{1: 1.2, 2:0.6},
{2: 0.6}]
T_alpha_total = [1.2,
2.4,
1.8,
0.6]
main_data_sample = mmvt_analyze.MMVT_data_sample(
model, N_alpha_beta, k_alpha_beta, N_i_j_alpha,
R_i_alpha_total, T_alpha_total)
main_data_sample.calculate_pi_alpha()
main_data_sample.fill_out_data_quantities()
main_data_sample.compute_rate_matrix()
main_data_sample.calculate_thermodynamics()
main_data_sample.calculate_kinetics()
mmvt_Q = main_data_sample.Q
# Elber stats
N_i_j = {(0,1): 4, (1,0): 4, (1,2): 2, (2,1): 2}
R_i = {0: 2.4, 1: 2.4, 2: 1.2}
elber_N = np.array([[0, 4, 0],
[4, 0, 2],
[0, 2, 0]])
elber_R = np.array([[2.4],
[2.4],
[1.2]])
elber_Q = np.zeros((n_milestones, n_milestones))
for i in range(n_milestones):
for j in range(n_milestones):
key = (i,j)
if key in N_i_j:
elber_Q[i,j] = N_i_j[key] / R_i[i]
for i in range(n_milestones):
elber_Q[i,i] = -np.sum(elber_Q[i,:])
# Make sure the two models make identical matrices
assert np.isclose(mmvt_Q, elber_Q).all()
# Now compare the distributions of both of them
# MMVT matrix sampler
data_sample_list, p_i_error, free_energy_profile_err, MFPTs_error, \
k_off_error, k_ons_error = mmvt_analyze.monte_carlo_milestoning_error(
main_data_sample, num=num, stride=stride, skip=skip, verbose=True)
mmvt_q1_distribution = []
mmvt_q2_distribution = []
for data_sample in data_sample_list:
mmvt_q1_distribution.append(data_sample.Q[0,1])
mmvt_q2_distribution.append(data_sample.Q[1,0])
# Elber matrix sampler
elber_q1_distribution = []
elber_q2_distribution = []
for counter in range(num * (stride) + skip):
#if verbose: print("MCMC stepnum: ", counter)
elber_Qnew = markov_chain_monte_carlo\
.irreversible_stochastic_matrix_algorithm_sample(
elber_Q, elber_N, elber_R)
if counter > skip and counter % stride == 0:
elber_q1_distribution.append(elber_Q[0,1])
elber_q2_distribution.append(elber_Q[1,0])
elber_Q = elber_Qnew
assert np.isclose(np.mean(elber_q1_distribution),
np.mean(mmvt_q1_distribution), rtol=0.5, atol=0.01)
assert np.isclose(np.std(elber_q1_distribution),
np.std(mmvt_q1_distribution), rtol=0.5, atol=0.01)
assert np.isclose(np.mean(elber_q2_distribution),
np.mean(mmvt_q2_distribution), rtol=0.5, atol=0.01)
assert np.isclose(np.std(elber_q2_distribution),
np.std(mmvt_q2_distribution), rtol=0.5, atol=0.01)
return
def test_MMVT_data_sample_fill_N_R_alpha_from_matrices():
model = make_simple_model()
N_i_j_alpha_exp = [{},
{(0,1):4, (1,0):4},
{(1,2):2, (2,1):2},
{}]
R_i_alpha_total_exp = [{0: 1.2},
{0: 1.2, 1:1.2},
{1: 1.2, 2:0.6},
{2: 0.6}]
mmvt_Nij_alpha = [
np.array([
[0, 0, 0],
[0, 0, 0],
[0, 0, 0]]),
np.array([
[0, 4, 0],
[4, 0, 0],
[0, 0, 0]]),
np.array([
[0, 0, 0],
[0, 0, 2],
[0, 2, 0]]),
np.array([
[0, 0, 0],
[0, 0, 0],
[0, 0, 0]])]
mmvt_Ri_alpha = [
np.array([
[1.2],
[0],
[0]]),
np.array([
[1.2],
[1.2],
[0]]),
np.array([
[0],
[1.2],
[0.6]]),
np.array([
[0],
[0],
[0.6]])]
N_i_j_alpha_start = [{},
{(0,1):3, (1,0):3},
{(1,2):3, (2,1):3},
{}]
R_i_alpha_total_start = [{0: 1.0},
{0: 1.0, 1:1.0},
{1: 1.0, 2:0.4},
{2: 0.4}]
T_alpha_total_start = [1.2,
2.4,
1.8,
0.6]
main_data_sample = mmvt_analyze.MMVT_data_sample(
model, None, None, N_i_j_alpha_start,
R_i_alpha_total_start, T_alpha_total_start)
main_data_sample.fill_N_R_alpha_from_matrices(mmvt_Nij_alpha, mmvt_Ri_alpha)
assert len(N_i_j_alpha_exp) == len(main_data_sample.N_i_j_alpha)
for alpha in range(len(N_i_j_alpha_exp)):
mmvt_Nij_exp = N_i_j_alpha_exp[alpha]
mmvt_Nij = main_data_sample.N_i_j_alpha[alpha]
compare_dicts(mmvt_Nij_exp, mmvt_Nij)
assert len(R_i_alpha_total_exp) == len(main_data_sample.R_i_alpha)
for alpha in range(len(R_i_alpha_total_exp)):
mmvt_Ri_exp = R_i_alpha_total_exp[alpha]
mmvt_Ri = main_data_sample.R_i_alpha[alpha]
compare_dicts(mmvt_Ri_exp, mmvt_Ri)
return
def test_MMVT_data_sample_make_mcmc_quantities():
model = make_simple_model()
N_alpha_beta = {(0,1):12, (1,0):12,
(1,2):12, (2,1):12,
(2,3):6, (3,2):6}
k_alpha_beta = {(0,1):20.0, (1,0):10.0,
(1,2):10.0, (2,1):(40.0/3.0),
(2,3):(20.0/3.0), (3,2):20.0}
N_i_j_alpha = [{},
{(0,1):4, (1,0):4},
{(1,2):2, (2,1):2},
{}]
R_i_alpha_total = [{0: 1.2},
{0: 1.2, 1:1.2},
{1: 1.2, 2:0.6},
{2: 0.6}]
T_alpha_total = [1.2,
2.4,
1.8,
0.6]
main_data_sample = mmvt_analyze.MMVT_data_sample(
model, N_alpha_beta, k_alpha_beta, N_i_j_alpha,
R_i_alpha_total, T_alpha_total)
main_data_sample.calculate_pi_alpha()
k_alpha_beta_matrix, N_alpha_beta_matrix, T_alpha_matrix,\
mmvt_Nij_alpha, mmvt_Ri_alpha, mmvt_Qij_alpha, T \
= main_data_sample.make_mcmc_quantities()
k_alpha_beta_matrix_exp = np.array([
[-20.0, 20.0, 0.0, 0.0],
[10.0, -20.0, 10.0, 0.0],
[0.0, (40.0/3.0), -(40.0/3.0)-(20.0/3.0), (20.0/3.0)],
[0.0, 0.0, 20.0, -20.0]])
N_alpha_beta_matrix_exp = np.array([
[0, 12, 0, 0],
[12, 0, 12, 0],
[0, 12, 0, 6],
[0, 0, 6, 0]])
T_alpha_matrix_exp = np.array([T_alpha_total]).T
mmvt_Nij_alpha_exp = [
np.array([
[0, 0, 0],
[0, 0, 0],
[0, 0, 0]]),
np.array([
[0, 4, 0],
[4, 0, 0],
[0, 0, 0]]),
np.array([
[0, 0, 0],
[0, 0, 2],
[0, 2, 0]]),
np.array([
[0, 0, 0],
[0, 0, 0],
[0, 0, 0]])]
mmvt_Ri_alpha_exp = [
np.array([
[1.2],
[0],
[0]]),
np.array([
[1.2],
[1.2],
[0]]),
np.array([
[0],
[1.2],
[0.6]]),
np.array([
[0],
[0],
[0.6]])]
mmvt_Qij_alpha_exp = [
np.array([
[0, 0, 0],
[0, 0, 0],
[0, 0, 0]]),
np.array([
[-4/1.2, 4/1.2, 0],
[4/1.2, -4/1.2, 0],
[0, 0, 0]]),
np.array([
[0, 0, 0],
[0, -2/1.2, 2/1.2],
[0, 2/0.6, -2/0.6]]),
np.array([
[0, 0, 0],
[0, 0, 0],
[0, 0, 0]])]
assert np.all(np.isclose(k_alpha_beta_matrix, k_alpha_beta_matrix_exp))
assert np.all(np.isclose(N_alpha_beta_matrix, N_alpha_beta_matrix_exp))
assert np.all(np.isclose(T_alpha_matrix, T_alpha_matrix_exp))
assert len(mmvt_Nij_alpha_exp) == len(mmvt_Nij_alpha)
for mmvt_Nij, mmvt_Nij_exp in zip(mmvt_Nij_alpha, mmvt_Nij_alpha_exp):
assert np.all(np.isclose(mmvt_Nij, mmvt_Nij_exp))
assert len(mmvt_Ri_alpha_exp) == len(mmvt_Ri_alpha)
for mmvt_Ri, mmvt_Ri_exp in zip(mmvt_Ri_alpha, mmvt_Ri_alpha_exp):
assert np.all(np.isclose(mmvt_Ri, mmvt_Ri_exp))
assert len(mmvt_Qij_alpha_exp) == len(mmvt_Qij_alpha)
for mmvt_Qij, mmvt_Qij_exp in zip(mmvt_Qij_alpha, mmvt_Qij_alpha_exp):
assert np.all(np.isclose(mmvt_Qij, mmvt_Qij_exp))
return