def setUp(self):
"""Store state of the rng"""
self.state = np.random.mtrand.get_state()
"""Reseed the rng to enforce 'deterministic' behavior"""
np.random.mtrand.seed(42)
"""Meta-stable birth-death chain"""
b = 2
q = np.zeros(7)
p = np.zeros(7)
q[1:] = 0.5
p[0:-1] = 0.5
q[2] = 1.0 - 10 ** (-b)
q[4] = 10 ** (-b)
p[2] = 10 ** (-b)
p[4] = 1.0 - 10 ** (-b)
bdc = BirthDeathChain(q, p)
P = bdc.transition_matrix()
self.dtraj = generate_traj(P, 10000, start=0)
self.tau = 1
"""Estimate MSM"""
self.C_MSM = cmatrix(self.dtraj, self.tau, sliding=True)
self.lcc_MSM = largest_connected_set(self.C_MSM)
self.Ccc_MSM = connected_cmatrix(self.C_MSM, lcc=self.lcc_MSM)
self.P_MSM = tmatrix(self.Ccc_MSM, reversible=True)
self.mu_MSM = statdist(self.P_MSM)
self.k = 3
self.ts = timescales(self.P_MSM, k=self.k, tau=self.tau)
def setUp(self):
"""Store state of the rng"""
self.state = np.random.mtrand.get_state()
"""Reseed the rng to enforce 'deterministic' behavior"""
np.random.mtrand.seed(42)
"""Meta-stable birth-death chain"""
b = 2
q = np.zeros(7)
p = np.zeros(7)
q[1:] = 0.5
p[0:-1] = 0.5
q[2] = 1.0 - 10**(-b)
q[4] = 10**(-b)
p[2] = 10**(-b)
p[4] = 1.0 - 10**(-b)
bdc = BirthDeathChain(q, p)
P = bdc.transition_matrix()
self.dtraj = generate_traj(P, 10000, start=0)
self.tau = 1
"""Estimate MSM"""
self.C_MSM = cmatrix(self.dtraj, self.tau, sliding=True)
self.lcc_MSM = largest_connected_set(self.C_MSM)
self.Ccc_MSM = connected_cmatrix(self.C_MSM, lcc=self.lcc_MSM)
self.P_MSM = tmatrix(self.Ccc_MSM, reversible=True)
self.mu_MSM = statdist(self.P_MSM)
self.k = 3
self.ts = timescales(self.P_MSM, k=self.k, tau=self.tau)
def compute(self):
"""Compute count matrix"""
self.C = cmatrix(self.dtrajs, self.lagtime, sliding=self.sliding)
"""Largest connected set"""
self.lcc = largest_connected_set(self.C)
"""Largest connected countmatrix"""
self.Ccc = connected_cmatrix(self.C)
"""Estimate transition matrix"""
self.T = tmatrix(self.Ccc, reversible=self.reversible)
self.computed=True
def _estimate_ts_tau(self, C, tau):
r"""Estimate timescales from the given count matrix.
"""
# connected set
C = (connected_cmatrix(C)).toarray()
if (len(C) > 1):
# estimate transition matrix
T = tmatrix(C, reversible=self._reversible)
# timescales
ts = timescales(T, tau, k=min(self._nits, len(T)) + 1)[1:]
return ts
else:
return None # no timescales available
def compute(self):
"""Compute count matrix"""
self.C = cmatrix(self.dtrajs, self.lagtime, sliding=self.sliding)
"""Largest connected set"""
self.lcc = largest_connected_set(self.C)
"""Largest connected countmatrix"""
self.Ccc = connected_cmatrix(self.C)
"""Estimate transition matrix"""
if self.estimate_on_lcc:
self.T = tmatrix(self.Ccc, reversible=self.reversible)
else:
self.T = tmatrix(self.C, reversible=self.reversible)
self.computed = True
def setUp(self):
"""Store state of the rng"""
self.state = np.random.mtrand.get_state()
"""Reseed the rng to enforce 'deterministic' behavior"""
np.random.mtrand.seed(42)
"""Meta-stable birth-death chain"""
b = 2
q = np.zeros(7)
p = np.zeros(7)
q[1:] = 0.5
p[0:-1] = 0.5
q[2] = 1.0 - 10**(-b)
q[4] = 10**(-b)
p[2] = 10**(-b)
p[4] = 1.0 - 10**(-b)
bdc = BirthDeathChain(q, p)
P = bdc.transition_matrix()
dtraj = generate_traj(P, 10000, start=0)
tau = 1
"""Estimate MSM"""
C_MSM = cmatrix(dtraj, tau)
lcc_MSM = largest_connected_set(C_MSM)
Ccc_MSM = connected_cmatrix(C_MSM, lcc=lcc_MSM)
P_MSM = tmatrix(Ccc_MSM)
mu_MSM = statdist(P_MSM)
"""Meta-stable sets"""
A = [0, 1, 2]
B = [4, 5, 6]
w_MSM = np.zeros((2, mu_MSM.shape[0]))
w_MSM[0, A] = mu_MSM[A] / mu_MSM[A].sum()
w_MSM[1, B] = mu_MSM[B] / mu_MSM[B].sum()
K = 10
P_MSM_dense = P_MSM.toarray()
p_MSM = np.zeros((K, 2))
w_MSM_k = 1.0 * w_MSM
for k in range(1, K):
w_MSM_k = np.dot(w_MSM_k, P_MSM_dense)
p_MSM[k, 0] = w_MSM_k[0, A].sum()
p_MSM[k, 1] = w_MSM_k[1, B].sum()
"""Assume that sets are equal, A(\tau)=A(k \tau) for all k"""
w_MD = 1.0 * w_MSM
p_MD = np.zeros((K, 2))
eps_MD = np.zeros((K, 2))
p_MSM[0, :] = 1.0
p_MD[0, :] = 1.0
eps_MD[0, :] = 0.0
for k in range(1, K):
"""Build MSM at lagtime k*tau"""
C_MD = cmatrix(dtraj, k * tau, sliding=True) / (k * tau)
lcc_MD = largest_connected_set(C_MD)
Ccc_MD = connected_cmatrix(C_MD, lcc=lcc_MD)
c_MD = Ccc_MD.sum(axis=1)
P_MD = tmatrix(Ccc_MD).toarray()
w_MD_k = np.dot(w_MD, P_MD)
"""Set A"""
prob_MD = w_MD_k[0, A].sum()
c = c_MD[A].sum()
p_MD[k, 0] = prob_MD
eps_MD[k, 0] = np.sqrt(k * (prob_MD - prob_MD**2) / c)
"""Set B"""
prob_MD = w_MD_k[1, B].sum()
c = c_MD[B].sum()
p_MD[k, 1] = prob_MD
eps_MD[k, 1] = np.sqrt(k * (prob_MD - prob_MD**2) / c)
"""Input"""
self.P_MSM = P_MSM
self.lcc_MSM = lcc_MSM
self.dtraj = dtraj
self.tau = tau
self.K = K
self.A = A
self.B = B
"""Expected results"""
self.p_MSM = p_MSM
self.p_MD = p_MD
self.eps_MD = eps_MD
def cktest(T_MSM,
lcc_MSM,
dtrajs,
lag,
K,
nsets=2,
sets=None,
full_output=False):
r"""Perform Chapman-Kolmogorov tests for given data.
Parameters
----------
T_MSM : (M, M) ndarray or scipy.sparse matrix
Transition matrix of estimated MSM
lcc_MSM : array-like
Largest connected set of the estimated MSM
dtrajs : list
discrete trajectories
lag : int
lagtime for the MSM estimation
K : int
number of time points for the test
nsets : int, optional
number of PCCA sets on which to perform the test
sets : list, optional
List of user defined sets for the test
full_output : bool, optional
Return additional information about set_factors
Returns
-------
p_MSM : (K, nsets) ndarray
p_MSM[k, l] is the probability of making a transition from
set l to set l after k*lag steps for the MSM computed at 1*lag
p_MD : (K, nsets) ndarray
p_MD[k, l] is the probability of making a transition from
set l to set l after k*lag steps as estimated from the given data
eps_MD : (K, nsets)
eps_MD[k, l] is an estimate for the statistical error of p_MD[k, l]
set_factors : (K, nsets) ndarray, optional
set_factor[k, i] is the quotient of the MD and the MSM set probabilities
References
----------
.. [1] Prinz, J H, H Wu, M Sarich, B Keller, M Senne, M Held, J D
Chodera, C Schuette and F Noe. 2011. Markov models of
molecular kinetics: Generation and validation. J Chem Phys
134: 174105
"""
p_MD = np.zeros((K, nsets))
p_MSM = np.zeros((K, nsets))
eps_MD = np.zeros((K, nsets))
set_factors = np.zeros((K, nsets))
if sets is None:
"""Compute PCCA-sets from MSM at lagtime 1*\tau"""
if issparse(T_MSM):
msg = (
"Converting sparse transition matrix to dense\n"
"since PCCA is currently only implemented for dense matrices.\n"
"You can avoid automatic conversion to dense arrays by\n"
"giving sets for the Chapman-Kolmogorov test explicitly")
warnings.warn(msg, UserWarning)
sets = pcca_sets(T_MSM.toarray(), nsets)
else:
sets = pcca_sets(T_MSM, nsets)
nsets = len(sets)
# translate sets from connected-set indexes to full indexes, where the comparison is made:
for i, s in enumerate(sets):
sets[i] = lcc_MSM[s]
"""Stationary distribution at 1*tau"""
mu_MSM = statdist(T_MSM)
"""Mapping to lcc at lagtime 1*tau"""
lccmap_MSM = MapToConnectedStateLabels(lcc_MSM)
"""Compute stationary distribution on sets"""
w_MSM_1 = np.zeros((nsets, mu_MSM.shape[0]))
for l in range(nsets):
A = sets[l]
A_MSM = lccmap_MSM.map(A)
w_MSM_1[l, A_MSM] = mu_MSM[A_MSM] / mu_MSM[A_MSM].sum()
w_MSM_k = 1.0 * w_MSM_1
p_MSM[0, :] = 1.0
p_MD[0, :] = 1.0
eps_MD[0, :] = 0.0
set_factors[0, :] = 1.0
for k in range(1, K):
"""Propagate probability vectors for MSM"""
w_MSM_k = propagate(w_MSM_k, T_MSM)
"""Estimate model at k*tau and normalize to make 'uncorrelated'"""
C_MD = cmatrix(dtrajs, k * lag, sliding=True) / (k * lag)
lcc_MD = largest_connected_set(C_MD)
Ccc_MD = connected_cmatrix(C_MD, lcc=lcc_MD)
"""State counts for MD"""
c_MD = Ccc_MD.sum(axis=1)
"""Transition matrix at k*tau"""
T_MD = tmatrix(Ccc_MD)
"""Mapping to lcc at lagtime k*tau"""
lccmap_MD = MapToConnectedStateLabels(lcc_MD)
"""Intersection of lcc_1 and lcc_k. lcc_k is not necessarily contained within lcc_1"""
lcc = np.intersect1d(lcc_MSM, lcc_MD)
"""Stationary distribution restricted to lcc at lagtime k*tau"""
mu_MD = np.zeros(T_MD.shape[0])
"""Extract stationary values in 'joint' lcc and assining them to their respective position"""
mu_MD[lccmap_MD.map(lcc)] = mu_MSM[lccmap_MSM.map(lcc)]
"""Obtain sets and distribution at k*tau"""
w_MD_1 = np.zeros((nsets, mu_MD.shape[0]))
for l in range(len(sets)):
A = sets[l]
"""Intersect the set with the lcc at lagtime k*tau"""
A_new = np.intersect1d(A, lcc)
if A_new.size > 0:
A_MD = lccmap_MD.map(A_new)
w_MD_1[l, A_MD] = mu_MD[A_MD] / mu_MD[A_MD].sum()
"""Propagate vector by the MD model"""
w_MD_k = propagate(w_MD_1, T_MD)
"""Compute values"""
for l in range(len(sets)):
A = sets[l]
"""MSM values"""
A_MSM = lccmap_MSM.map(A)
p_MSM[k, l] = w_MSM_k[l, A_MSM].sum()
"""MD values"""
A_new = np.intersect1d(A, lcc)
if A_new.size > 0:
A_MD = lccmap_MD.map(A_new)
prob_MD = w_MD_k[l, A_MD].sum()
p_MD[k, l] = prob_MD
"""Statistical errors"""
c = c_MD[A_MD].sum()
eps_MD[k, l] = np.sqrt(k * (prob_MD - prob_MD**2) / c)
set_factors[k, l] = mu_MSM[lccmap_MSM.map(
A_new)].sum() / mu_MSM[A_MSM].sum()
if full_output:
return p_MSM, p_MD, eps_MD, set_factors
else:
return p_MSM, p_MD, eps_MD
def setUp(self):
"""Store state of the rng"""
self.state = np.random.mtrand.get_state()
"""Reseed the rng to enforce 'deterministic' behavior"""
np.random.mtrand.seed(42)
"""Meta-stable birth-death chain"""
b = 2
q = np.zeros(7)
p = np.zeros(7)
q[1:] = 0.5
p[0:-1] = 0.5
q[2] = 1.0 - 10 ** (-b)
q[4] = 10 ** (-b)
p[2] = 10 ** (-b)
p[4] = 1.0 - 10 ** (-b)
bdc = BirthDeathChain(q, p)
P = bdc.transition_matrix()
dtraj = generate_traj(P, 10000, start=0)
tau = 1
"""Estimate MSM"""
MSM = estimate_markov_model(dtraj, tau)
C_MSM = MSM.count_matrix_full
lcc_MSM = MSM.largest_connected_set
Ccc_MSM = MSM.count_matrix_active
P_MSM = MSM.transition_matrix
mu_MSM = MSM.stationary_distribution
"""Meta-stable sets"""
A = [0, 1, 2]
B = [4, 5, 6]
w_MSM = np.zeros((2, mu_MSM.shape[0]))
w_MSM[0, A] = mu_MSM[A] / mu_MSM[A].sum()
w_MSM[1, B] = mu_MSM[B] / mu_MSM[B].sum()
K = 10
P_MSM_dense = P_MSM
p_MSM = np.zeros((K, 2))
w_MSM_k = 1.0 * w_MSM
for k in range(1, K):
w_MSM_k = np.dot(w_MSM_k, P_MSM_dense)
p_MSM[k, 0] = w_MSM_k[0, A].sum()
p_MSM[k, 1] = w_MSM_k[1, B].sum()
"""Assume that sets are equal, A(\tau)=A(k \tau) for all k"""
w_MD = 1.0 * w_MSM
p_MD = np.zeros((K, 2))
eps_MD = np.zeros((K, 2))
p_MSM[0, :] = 1.0
p_MD[0, :] = 1.0
eps_MD[0, :] = 0.0
for k in range(1, K):
"""Build MSM at lagtime k*tau"""
C_MD = cmatrix(dtraj, k * tau, sliding=True) / (k * tau)
lcc_MD = largest_connected_set(C_MD)
Ccc_MD = connected_cmatrix(C_MD, lcc=lcc_MD)
c_MD = Ccc_MD.sum(axis=1)
P_MD = tmatrix(Ccc_MD).toarray()
w_MD_k = np.dot(w_MD, P_MD)
"""Set A"""
prob_MD = w_MD_k[0, A].sum()
c = c_MD[A].sum()
p_MD[k, 0] = prob_MD
eps_MD[k, 0] = np.sqrt(k * (prob_MD - prob_MD ** 2) / c)
"""Set B"""
prob_MD = w_MD_k[1, B].sum()
c = c_MD[B].sum()
p_MD[k, 1] = prob_MD
eps_MD[k, 1] = np.sqrt(k * (prob_MD - prob_MD ** 2) / c)
"""Input"""
self.MSM = MSM
self.K = K
self.A = A
self.B = B
"""Expected results"""
self.p_MSM = p_MSM
self.p_MD = p_MD
self.eps_MD = eps_MD
def cktest(T_MSM, lcc_MSM, dtrajs, lag, K, nsets=2, sets=None, full_output=False):
r"""Perform Chapman-Kolmogorov tests for given data.
Parameters
----------
T_MSM : (M, M) ndarray or scipy.sparse matrix
Transition matrix of estimated MSM
lcc_MSM : array-like
Largest connected set of the estimated MSM
dtrajs : list
discrete trajectories
lag : int
lagtime for the MSM estimation
K : int
number of time points for the test
nsets : int, optional
number of PCCA sets on which to perform the test
sets : list, optional
List of user defined sets for the test
full_output : bool, optional
Return additional information about set_factors
Returns
-------
p_MSM : (K, nsets) ndarray
p_MSM[k, l] is the probability of making a transition from
set l to set l after k*lag steps for the MSM computed at 1*lag
p_MD : (K, nsets) ndarray
p_MD[k, l] is the probability of making a transition from
set l to set l after k*lag steps as estimated from the given data
eps_MD : (K, nsets)
eps_MD[k, l] is an estimate for the statistical error of p_MD[k, l]
set_factors : (K, nsets) ndarray, optional
set_factor[k, i] is the quotient of the MD and the MSM set probabilities
References
----------
.. [1] Prinz, J H, H Wu, M Sarich, B Keller, M Senne, M Held, J D
Chodera, C Schuette and F Noe. 2011. Markov models of
molecular kinetics: Generation and validation. J Chem Phys
134: 174105
"""
p_MD = np.zeros((K, nsets))
p_MSM = np.zeros((K, nsets))
eps_MD = np.zeros((K, nsets))
set_factors = np.zeros((K, nsets))
if sets is None:
"""Compute PCCA-sets from MSM at lagtime 1*\tau"""
if issparse(T_MSM):
msg = ("Converting sparse transition matrix to dense\n"
"since PCCA is currently only implemented for dense matrices.\n"
"You can avoid automatic conversion to dense arrays by\n"
"giving sets for the Chapman-Kolmogorov test explicitly")
warnings.warn(msg, UserWarning)
sets = pcca_sets(T_MSM.toarray(), nsets)
else:
sets = pcca_sets(T_MSM, nsets)
nsets = len(sets)
# translate sets from connected-set indexes to full indexes, where the comparison is made:
for i, s in enumerate(sets):
sets[i] = lcc_MSM[s]
"""Stationary distribution at 1*tau"""
mu_MSM = statdist(T_MSM)
"""Mapping to lcc at lagtime 1*tau"""
lccmap_MSM = MapToConnectedStateLabels(lcc_MSM)
"""Compute stationary distribution on sets"""
w_MSM_1 = np.zeros((nsets, mu_MSM.shape[0]))
for l in range(nsets):
A = sets[l]
A_MSM = lccmap_MSM.map(A)
w_MSM_1[l, A_MSM] = mu_MSM[A_MSM] / mu_MSM[A_MSM].sum()
w_MSM_k = 1.0 * w_MSM_1
p_MSM[0, :] = 1.0
p_MD[0, :] = 1.0
eps_MD[0, :] = 0.0
set_factors[0, :] = 1.0
for k in range(1, K):
"""Propagate probability vectors for MSM"""
w_MSM_k = propagate(w_MSM_k, T_MSM)
"""Estimate model at k*tau and normalize to make 'uncorrelated'"""
C_MD = cmatrix(dtrajs, k * lag, sliding=True) / (k * lag)
lcc_MD = largest_connected_set(C_MD)
Ccc_MD = connected_cmatrix(C_MD, lcc=lcc_MD)
"""State counts for MD"""
c_MD = Ccc_MD.sum(axis=1)
"""Transition matrix at k*tau"""
T_MD = tmatrix(Ccc_MD)
"""Mapping to lcc at lagtime k*tau"""
lccmap_MD = MapToConnectedStateLabels(lcc_MD)
"""Intersection of lcc_1 and lcc_k. lcc_k is not necessarily contained within lcc_1"""
lcc = np.intersect1d(lcc_MSM, lcc_MD)
"""Stationary distribution restricted to lcc at lagtime k*tau"""
mu_MD = np.zeros(T_MD.shape[0])
"""Extract stationary values in 'joint' lcc and assining them to their respective position"""
mu_MD[lccmap_MD.map(lcc)] = mu_MSM[lccmap_MSM.map(lcc)]
"""Obtain sets and distribution at k*tau"""
w_MD_1 = np.zeros((nsets, mu_MD.shape[0]))
for l in range(len(sets)):
A = sets[l]
"""Intersect the set with the lcc at lagtime k*tau"""
A_new = np.intersect1d(A, lcc)
if A_new.size > 0:
A_MD = lccmap_MD.map(A_new)
w_MD_1[l, A_MD] = mu_MD[A_MD] / mu_MD[A_MD].sum()
"""Propagate vector by the MD model"""
w_MD_k = propagate(w_MD_1, T_MD)
"""Compute values"""
for l in range(len(sets)):
A = sets[l]
"""MSM values"""
A_MSM = lccmap_MSM.map(A)
p_MSM[k, l] = w_MSM_k[l, A_MSM].sum()
"""MD values"""
A_new = np.intersect1d(A, lcc)
if A_new.size > 0:
A_MD = lccmap_MD.map(A_new)
prob_MD = w_MD_k[l, A_MD].sum()
p_MD[k, l] = prob_MD
"""Statistical errors"""
c = c_MD[A_MD].sum()
eps_MD[k, l] = np.sqrt(k * (prob_MD - prob_MD ** 2) / c)
set_factors[k, l] = mu_MSM[lccmap_MSM.map(A_new)].sum() / mu_MSM[A_MSM].sum()
if full_output:
return p_MSM, p_MD, eps_MD, set_factors
else:
return p_MSM, p_MD, eps_MD
def chapman_kolmogorov(dtrajs, lag, K, nsets=2, sets=None):
r"""Perform Chapman-Kolmogorov tests for given data.
Parameters
----------
dtrajs : list
discrete trajectories
lag : int
lagtime for the MSM estimation
K : int
number of time points for the test
nsets : int, optional
number of PCCA sets on which to perform the test
sets : list, optional
List of user defined sets for the test
Returns
-------
p_MSM : (K, n_sets) ndarray
p_MSM[k, l] is the probability of making a transition from
set l to set l after k*lag steps for the MSM computed at 1*lag
p_MD : (K, n_sets) ndarray
p_MD[k, l] is the probability of making a transition from
set l to set l after k*lag steps as estimated from the given data
eps_MD : (K, n_sets)
eps_MD[k, l] is an estimate for the statistical error of p_MD[k, l]
References
----------
.. [1] Prinz, J H, H Wu, M Sarich, B Keller, M Senne, M Held, J D
Chodera, C Schuette and F Noe. 2011. Markov models of
molecular kinetics: Generation and validation. J Chem Phys
134: 174105
"""
C_1=cmatrix(dtrajs, lag, sliding=True)
lcc_1=largest_connected_set(C_1)
"""Compute PCCA-sets from MSM at lagtime 1*\tau"""
if sets is None:
Ccc_1=connected_cmatrix(C_1, lcc=lcc_1)
T_1=tmatrix(Ccc_1)
sets=pcca_sets(T_1.toarray(), nsets, lcc_1)
p_MD = np.zeros((K, nsets))
p_MSM = np.zeros((K, nsets))
eps_MD = np.zeros((K, nsets))
for k in range(1, K):
C_k = cmatrix(dtrajs, (k+1)*lag, sliding=True)
lcc_k = largest_connected_set(C_k)
lcc = np.intersect1d(lcc_1, lcc_k)
Ccc_1 = connected_cmatrix(C_1, lcc=lcc)
Ccc_k = connected_cmatrix(C_k, lcc=lcc)
T_1 = tmatrix(Ccc_1).toarray()
T_k = tmatrix(Ccc_k).toarray()
mu = statdist(T_1)
T_1_k = np.linalg.matrix_power(T_1, k+1)
lccmap=MapToConnectedStateLabels(lcc)
C=Ccc_k.toarray()
for l in range(len(sets)):
w=np.zeros(len(lcc))
inds = lccmap.map(sets[l])
nu = mu[inds]
nu /= nu.sum()
w[inds] = nu
w_1_k = np.dot(w, T_1_k)
w_k = np.dot(w, T_k)
prob_MD=np.sum(w_k[inds])
prob_MSM=np.sum(w_1_k[inds])
p_MD[k, l] = prob_MD
p_MSM[k, l] = prob_MSM
"""Statistical errors"""
c=C[inds, :].sum()
eps_MD[k, l]=np.sqrt((k + 1) * (prob_MD - prob_MD**2) / c)
return p_MSM, p_MD, eps_MD