def __init__(self, system, platform=None):
"""
Initialize a transition path sampling simulation with activity field strength s.
ARGUMENTS
system (simtk.chem.openSystem) - the molecular mechanics system
N (int) - number of particles in the system
NA (int) - number of A-type particles to be used in computing activity functional K[x(t)]
mass (simtk.unit.Quantity with units mass/particle) - particle mass to be used in computing reduced quantites
epsilon (simtk.unit.Quantity with units energy/mole) - Lennard-Jones well depth parameter to be used in computing reduced quantities
sigma (simtk.unit.Quantity with units distance) - Lennard-Jones sigma parameter to be used in computing reduced quantities
timestep (simtk.unit.Quantity with units time) - timestep to be used in dynamics simulations
nsteps_per_frame (int) - number of MD timesteps per simulation snapshot or 'frame' in a trajectory
nframes (int) - number of simulation snapshots or 'frames' for a fixed-length TPS trajectory, not counting initial configuration
temperture (simtk.unit.Quantity with units temperature) - simulation temeprature
s (float) - the value of the field parameter
OPTIONAL ARGUMENTS
platform (simtk.chem.openPlatform) - platform to use for OpenMM pathsimulators
"""
# Store local copy of System.
self.system = system
# self.mass = mass
# self.epsilon = epsilon
# self.sigma = sigma
#
# self.timestep = timestep
# self.nsteps_per_frame = nsteps_per_frame
# self.delta_t = timestep * nsteps_per_frame
#
# Compute thermal energy and inverse temperature from specified temperature.
self.kT = kB * self.temperature # thermal energy
self.beta = 1.0 / self.kT # inverse temperature
# Create a Context for integration.
if platform:
self.context = Context(self.system, self.integrator, platform)
else:
self.context = Context(self.system, self.integrator)
# Store reduced units
# self.t_obs = nframes * self.delta_t
# self.s_reduced_unit = 1.0 / (self.sigma**2 * self.delta_t)
# self.K_reduced_unit = (self.N * self.t_obs * self.sigma**2)
# self.H_reduced_unit = self.epsilon # reduced unit for path Hamiltonian (energy)
# self.beta_reduced_unit = 1.0 / self.epsilon # reduced unit for inverse temperature
self.temperature = 298.0 * kelvin
self.collision_rate = 91.0 / picoseconds
self.timestep = 1.0 * femtoseconds
self.integrator = VVVRIntegrator(self.temperature, self.collision_rate, self.timestep)
return
pressure = 1.0 * unit.atmospheres
temperature = 300.0 * unit.kelvin
temperature_hot = 600.0 * unit.kelvin
frequency = 25
# If system is periodic, equilibrate at 1 atm at 300 K.
if system.usesPeriodicBoundaryConditions():
print('Equilibrating with barostat...')
import copy
barostat = openmm.MonteCarloBarostat(pressure, temperature, frequency)
system_with_barostat = copy.deepcopy(system)
system_with_barostat.addForce(barostat)
collision_rate = 10.0 / unit.picoseconds
timestep = 2.0 * unit.femtoseconds
#integrator = openmm.LangevinIntegrator(temperature, collision_rate, timestep)
integrator = VVVRIntegrator(temperature, collision_rate, timestep)
context = openmm.Context(system_with_barostat, integrator)
context.setPositions(positions)
niterations = 20
nsteps = 500
for iteration in range(niterations):
print('Iteration %5d / %5d: volume = %8.3f nm^3' % (iteration, niterations, context.getState().getPeriodicBoxVolume() / unit.nanometers**3))
integrator.step(nsteps)
positions = context.getState(getPositions=True).getPositions(asNumpy=True)
box_vectors = context.getState().getPeriodicBoxVectors()
system.setDefaultPeriodicBoxVectors(*box_vectors)
del context, integrator, system_with_barostat
else:
print('System does not use periodic boundary conditions; skipping equilibration.')
# Create test system
pressure = 1.0 * unit.atmospheres
temperature = 300.0 * unit.kelvin
temperature_hot = 600.0 * unit.kelvin
frequency = 25
# If system is periodic, equilibrate at 1 atm at 300 K.
if system.usesPeriodicBoundaryConditions():
print('Equilibrating with barostat...')
import copy
barostat = openmm.MonteCarloBarostat(pressure, temperature, frequency)
system_with_barostat = copy.deepcopy(system)
system_with_barostat.addForce(barostat)
collision_rate = 10.0 / unit.picoseconds
timestep = 2.0 * unit.femtoseconds
#integrator = openmm.LangevinIntegrator(temperature, collision_rate, timestep)
integrator = VVVRIntegrator(temperature, collision_rate, timestep)
context = openmm.Context(system_with_barostat, integrator)
context.setPositions(positions)
niterations = 20
nsteps = 500
for iteration in range(niterations):
print('Iteration %5d / %5d: volume = %8.3f nm^3' %
(iteration, niterations,
context.getState().getPeriodicBoxVolume() / unit.nanometers**3))
integrator.step(nsteps)
positions = context.getState(getPositions=True).getPositions(asNumpy=True)
box_vectors = context.getState().getPeriodicBoxVectors()
system.setDefaultPeriodicBoxVectors(*box_vectors)
del context, integrator, system_with_barostat
else:
print(
class TransitionPathSampling(object):
"""
General Transition path sampling with stopping criterion
"""
def __init__(self, system, platform=None):
"""
Initialize a transition path sampling simulation with activity field strength s.
ARGUMENTS
system (simtk.chem.openSystem) - the molecular mechanics system
N (int) - number of particles in the system
NA (int) - number of A-type particles to be used in computing activity functional K[x(t)]
mass (simtk.unit.Quantity with units mass/particle) - particle mass to be used in computing reduced quantites
epsilon (simtk.unit.Quantity with units energy/mole) - Lennard-Jones well depth parameter to be used in computing reduced quantities
sigma (simtk.unit.Quantity with units distance) - Lennard-Jones sigma parameter to be used in computing reduced quantities
timestep (simtk.unit.Quantity with units time) - timestep to be used in dynamics simulations
nsteps_per_frame (int) - number of MD timesteps per simulation snapshot or 'frame' in a trajectory
nframes (int) - number of simulation snapshots or 'frames' for a fixed-length TPS trajectory, not counting initial configuration
temperture (simtk.unit.Quantity with units temperature) - simulation temeprature
s (float) - the value of the field parameter
OPTIONAL ARGUMENTS
platform (simtk.chem.openPlatform) - platform to use for OpenMM pathsimulators
"""
# Store local copy of System.
self.system = system
# self.mass = mass
# self.epsilon = epsilon
# self.sigma = sigma
#
# self.timestep = timestep
# self.nsteps_per_frame = nsteps_per_frame
# self.delta_t = timestep * nsteps_per_frame
#
# Compute thermal energy and inverse temperature from specified temperature.
self.kT = kB * self.temperature # thermal energy
self.beta = 1.0 / self.kT # inverse temperature
# Create a Context for integration.
if platform:
self.context = Context(self.system, self.integrator, platform)
else:
self.context = Context(self.system, self.integrator)
# Store reduced units
# self.t_obs = nframes * self.delta_t
# self.s_reduced_unit = 1.0 / (self.sigma**2 * self.delta_t)
# self.K_reduced_unit = (self.N * self.t_obs * self.sigma**2)
# self.H_reduced_unit = self.epsilon # reduced unit for path Hamiltonian (energy)
# self.beta_reduced_unit = 1.0 / self.epsilon # reduced unit for inverse temperature
self.temperature = 298.0 * kelvin
self.collision_rate = 91.0 / picoseconds
self.timestep = 1.0 * femtoseconds
self.integrator = VVVRIntegrator(self.temperature, self.collision_rate, self.timestep)
return
def generateTrajectory(self, x0, nframes):
"""
Generate a velocity Verlet trajectory consisting of ntau segments of tau_steps in between storage of Snapshots and randomization of velocities.
ARGUMENTS
x0 (coordinate set) - initial coordinates; velocities will be assigned from Maxwell-Boltzmann distribution
nframes (int) - number of trajectory segments to generate
RETURNS
trajectory (list of Snapshot) - generated trajectory of initial conditions, including initial coordinate set
NOTES
This exists in OpenMM. Check with John on how to best get snapshots
This routine generates a velocity Verlet trajectory for systems without constraints by wrapping the OpenMM 'VerletIntegrator' in two half-kicks of the velocity.
"""
# Set initial positions
self.context.setPositions(x0)
# Store initial state for each trajectory segment in trajectory.
trajectory = Trajectory()
# Construct mass vector.
nparticles = self.system.getNumParticles()
mass = Quantity(numpy.zeros([nparticles, 3], numpy.float64), amu)
for particle_index in range(nparticles):
mass[particle_index, :] = self.system.getParticleMass(particle_index)
# Assign velocities from Maxwell-Boltzmann distribution
self.context.setVelocitiesToTemperature(self.temperature)
# Store initial snapshot of trajectory segment.
snapshot = Snapshot(context=self.context)
trajectory.forward(snapshot)
# Propagate dynamics by velocity Verlet.
in_void = True
self.frame = 0
while self.frame < self.xframes and in_void == True:
# Do integrator x steps
self.integrator.step()
# Check if reached a core set
in_void = self.check_void()
# Store final snapshot of trajectory.
snapshot = Snapshot(self.context)
trajectory.forward(snapshot)
return trajectory
def logEquilibriumTrajectoryProbability(self, trajectory):
"""
Compute the log equilibrium probability (up to an unknown additive constant) of an unbiased trajectory evolved according to Verlet dynamics with Andersen thermostatting.
ARGUMENTS
trajectory (Trajectory) - the trajectory
RETURNS
log_q (float) - the log equilibrium probability of the trajectory
NOTES
This might be better places into the trajectory class. The trajectory should know the system and ensemble? and so it is not necessarily
TPS specific
"""
nsnapshots = len(trajectory)
log_q = -self.beta * trajectory[0].total_energy
for snapshot_index in range(1, nsnapshots - 1):
log_q += -self.beta * trajectory[snapshot_index].kinetic_energy
return log_q
def pathHamiltonian(self, trajectory):
"""
Compute the generalized path Hamiltonian of the trajectory.
ARGUMENTS
trajectory (Trajectory) - the trajectory
RETURNS
H (simtk.unit.Quantity with units of energy) - the generalized path Hamiltonian
REFERENCES
For a description of the path Hamiltonian, see [1]:
[1] Chodera JD, Swope WC, Noe F, Prinz JH, Shirts MR, and Pande VS. Dynamical reweighting:
Improved estimates of dynamical properties from simulations at multiple temperatures.
NOTES
Could also more efficiently placed in the trajectory class
"""
nsnapshots = len(trajectory)
H = trajectory[0].total_energy
for snapshot_index in range(1, nsnapshots - 1):
H += trajectory[snapshot_index].kinetic_energy
return H
def computeActivity(self, trajectory):
"""
Compute the activity of a given trajectory, defined in Ref. [1] as
K[x(t)] = delta_t \sum_{t=0}^{t_obs} \sum_{j=1}^N [r_j(t+delta_t) - r_j(t)]^2
RETURNS
K (simtk.unit) - activity K[x(t)] for the specified trajectory
NOTES
Place also into trajectory class
"""
# Determine number of frames in trajectory.
nframes = len(trajectory)
# Compute activity of component A.
K = 0.0 * self.delta_t * nanometers ** 2
for frame_index in range(nframes - 1):
# Compute displacement of all atoms.
delta_r = trajectory[frame_index + 1].coordinates - trajectory[frame_index].coordinates
# Compute contribution to activity K.
K += self.delta_t * ((delta_r[0 : self.N, :] / nanometers) ** 2).sum() * (nanometers ** 2)
return K
def sampleTrajectory(self, trajectory):
"""
Conduct one step of transition path sampling MCMC to generate a (correlated) trajectory sample.
ARGUMENTS
trajectory (Trajectory) - a previous trajectory sample from the TPS ensemble
RETURN VALUES
trajectory (Trajectory) - new sampled trajectory (correlated with previous trajectory sample)
NOTE
The new trajectory may share Snapshot objects from the old trajectory; modification of these objects
will result in both old and new trajectories being updated. Make a deep copy if necessary to keep these
objects fully independent.
Snapshot objects should be immutable!!!! At least where we can make sure of this.
This should be changed to TIS, but should maybe keep all the trajetory idea, etc and just extend to TIS
"""
# Determine length of trajectory
nframes = len(trajectory)
# Compute value of activity K[x(t)]
K_old = self.computeActivity(trajectory)
# Choose a shooting or shift move
SHOOT_PROBABILITY = 0.5 # probability of picking a shooting move or a shift move
if numpy.random.rand() < SHOOT_PROBABILITY:
# Shoot part of a new trajectory
# Pick a timeslice to shoot from.
# TODO: This could be changed to more transition regions
frame_index = numpy.random.random_integers(1, nframes - 2)
# Pick a shooting direction.
if numpy.random.rand() < 0.5:
# Shoot forward.
print "Shooting forward from frame %d" % frame_index
partial_trajectory = self.generateTrajectory(
trajectory[frame_index].coordinates, nframes - frame_index - 1
)
trial_trajectory = trajectory[0:frame_index] + partial_trajectory
else:
# Shoot backwards
print "Shooting backward from frame %d" % frame_index
partial_trajectory = self.generateTrajectory(trajectory[frame_index].coordinates, frame_index)
partial_trajectory.reverse()
trial_trajectory = partial_trajectory[:-1] + trajectory[frame_index:]
else:
# Shift trajectory.
# Pick a timeslice to form start of new trajectory.
# Don't allow shifting by zero -- this screws with python indexing.
nshift = numpy.random.random_integers(1, nframes - 2)
# Pick a shooting direction.
if numpy.random.rand() < 0.5:
print "Shifting by +%d" % nshift
# Shoot forward from end.
partial_trajectory = self.generateTrajectory(trajectory[-1].coordinates, nshift)
trial_trajectory = trajectory[nshift:-1] + partial_trajectory
else:
# Shoot backwards from beginning.
print "Shifting by -%d" % nshift
partial_trajectory = self.generateTrajectory(trajectory[0].coordinates, nshift)
partial_trajectory.reverse()
trial_trajectory = partial_trajectory[:-1] + trajectory[0:-nshift]
# Compute new activity
K_trial = self.computeActivity(trial_trajectory)
# Accept or reject according to s-field.
# print "s * (sigma**2 * delta_t) = %.5f" % (self.s / self.s_reduced_unit)
# print "K_old / (N * t_obs * sigma**2) = %.5f" % (K_old / self.K_reduced_unit)
# print "K_trial / (N * t_obs * sigma**2) = %.5f" % (K_trial / self.K_reduced_unit)
log_P_accept = -self.s * (K_trial - K_old)
# print "log_P_accept = %f" % log_P_accept
self.nattempted += 1
if (log_P_accept > 0.0) or (numpy.random.rand() < math.exp(log_P_accept)):
# Accept trajectory move.
print "Accepted."
self.naccepted += 1
K_old = K_trial
trajectory = trial_trajectory
else:
# Move was rejected
print "Rejected."
pass
return trajectory
# In[2]:
# this cell is all OpenMM specific
forcefield = app.ForceField('amber96.xml', 'tip3p.xml')
pdb = app.PDBFile("AD_initial_frame.pdb")
system = forcefield.createSystem(
pdb.topology,
nonbondedMethod=app.PME,
nonbondedCutoff=1.0*unit.nanometers,
constraints=app.HBonds,
rigidWater=True,
ewaldErrorTolerance=0.0005
)
hi_T_integrator = VVVRIntegrator(
500*unit.kelvin,
1.0/unit.picoseconds,
2.0*unit.femtoseconds)
hi_T_integrator.setConstraintTolerance(0.00001)
# The storage file will need a template snapshot. In addition, the OPS OpenMM-based `Engine` has a few properties and options that are set by these dictionaries.
# In[3]:
template = omm.snapshot_from_pdb("AD_initial_frame.pdb")
openmm_properties = {'OpenCLPrecision': 'mixed'}
engine_options = {
'n_steps_per_frame': 10,
'n_frames_max': 2000
}
