def test_write_object():
"""Test writing of a object.
"""
tmpfile = tempfile.NamedTemporaryFile()
storage = NetCDFStorage(tmpfile.name, mode='w')
#use names we might encounter in simulation
envname = 'vacuum'
modname = 'ExpandedEnsembleSampler'
varname = 'energy'
view = NetCDFStorageView(storage, envname, modname)
obj = { 0 : 0 }
view.write_object('singleton', obj)
for iteration in range(10):
obj = { 'iteration' : iteration }
view.write_object(varname, obj, iteration=iteration)
for iteration in range(10):
obj = storage.get_object(envname, modname, varname, iteration=iteration)
assert ('iteration' in obj)
assert (obj['iteration'] == iteration)
def __init__(self, complex_sampler, solvent_sampler, log_state_penalties, storage=None, verbose=False):
"""
Initialize a protonation state sampler with fixed target probabilities for ligand in solvent.
Parameters
----------
complex_sampler : ExpandedEnsembleSampler
Ligand in complex sampler
solvent_sampler : SAMSSampler
Ligand in solution sampler
log_state_penalties : dict
log_state_penalties[smiles] is the log state free energy (in kT) for ligand state 'smiles'
storage : NetCDFStorage, optional, default=None
If specified, will use the storage layer to write trajectory data.
verbose : bool, optional, default=False
If true, will print verbose output
"""
# Store target samplers.
self.log_state_penalties = log_state_penalties
self.samplers = [complex_sampler, solvent_sampler]
self.complex_sampler = complex_sampler
self.solvent_sampler = solvent_sampler
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize storage for target probabilities.
self.log_target_probabilities = { key : - log_state_penalties[key] for key in log_state_penalties }
self.verbose = verbose
self.iteration = 0
def test_storage_view():
"""Test writing of a quantity.
"""
tmpfile = tempfile.NamedTemporaryFile()
storage = NetCDFStorage(tmpfile.name, mode='w')
view1 = NetCDFStorageView(storage, envname='envname')
view2 = NetCDFStorageView(view1, modname='modname')
assert (view1._envname == 'envname')
assert (view2._envname == 'envname')
assert (view2._modname == 'modname')
def test_write_quantity():
"""Test writing of a quantity.
"""
tmpfile = tempfile.NamedTemporaryFile()
storage = NetCDFStorage(tmpfile.name, mode='w')
view = NetCDFStorageView(storage, 'envname', 'modname')
view.write_quantity('singleton', 1.0)
for iteration in range(10):
view.write_quantity('varname', float(iteration), iteration=iteration)
for iteration in range(10):
assert (storage._ncfile['/envname/modname/varname'][iteration] == float(iteration))
def __init__(self, complex_sampler, solvent_sampler, log_state_penalties, storage=None, verbose=False):
"""
Initialize a protonation state sampler with fixed target probabilities for ligand in solvent.
Parameters
----------
complex_sampler : ExpandedEnsembleSampler
Ligand in complex sampler
solvent_sampler : SAMSSampler
Ligand in solution sampler
log_state_penalties : dict
log_state_penalties[smiles] is the log state free energy (in kT) for ligand state 'smiles'
storage : NetCDFStorage, optional, default=None
If specified, will use the storage layer to write trajectory data.
verbose : bool, optional, default=False
If true, will print verbose output
"""
# Store target samplers.
self.log_state_penalties = log_state_penalties
self.samplers = [complex_sampler, solvent_sampler]
self.complex_sampler = complex_sampler
self.solvent_sampler = solvent_sampler
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize storage for target probabilities.
self.log_target_probabilities = { key : - log_state_penalties[key] for key in log_state_penalties }
self.verbose = verbose
self.iteration = 0
def __init__(self, target_samplers, storage=None, verbose=False):
"""
Initialize a multi-objective design sampler with the specified target sampler powers.
Parameters
----------
target_samplers : dict
target_samplers[sampler] is the exponent associated with SAMS sampler `sampler` in the multi-objective design.
storage : NetCDFStorage, optional, default=None
If specified, will use the storage layer to write trajectory data.
verbose : bool, optional, default=False
If true, will print verbose output
The target sampler weights for N samplers with specified exponents \alpha_n are given by
\pi_{nk} \propto \prod_{n=1}^N Z_{nk}^{alpha_n}
where \pi_{nk} is the target weight for sampler n state k,
and Z_{nk} is the relative partition function of sampler n among states k.
Examples
--------
Set up a mutation sampler to maximize implicit solvent hydration free energy.
>>> from perses.tests.testsystems import AlanineDipeptideTestSystem
>>> testsystem = AlanineDipeptideTestSystem()
>>> # Set up target samplers.
>>> target_samplers = { testsystem.sams_samplers['implicit'] : 1.0, testsystem.sams_samplers['vacuum'] : -1.0 }
>>> # Set up the design sampler.
>>> designer = MultiTargetDesign(target_samplers)
"""
# Store target samplers.
self.sampler_exponents = target_samplers
self.samplers = list(target_samplers.keys())
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage,
modname=self.__class__.__name__)
# Initialize storage for target probabilities.
self.log_target_probabilities = dict()
self.verbose = verbose
self.iteration = 0
def test_write_object():
"""Test writing of a object.
"""
tmpfile = tempfile.NamedTemporaryFile()
storage = NetCDFStorage(tmpfile.name, mode='w')
view = NetCDFStorageView(storage, 'envname', 'modname')
obj = { 0 : 0 }
view.write_object('singleton', obj)
for iteration in range(10):
obj = { 'iteration' : iteration }
view.write_object('varname', obj, iteration=iteration)
for iteration in range(10):
encoded = storage._ncfile['/envname/modname/varname'][iteration]
obj = json.loads(encoded)
assert ('iteration' in obj)
assert (obj['iteration'] == iteration)
def test_write_object():
"""Test writing of a object.
"""
tmpfile = tempfile.NamedTemporaryFile()
storage = NetCDFStorage(tmpfile.name, mode='w')
#use names we might encounter in simulation
envname = 'vacuum'
modname = 'ExpandedEnsembleSampler'
varname = 'energy'
view = NetCDFStorageView(storage, envname, modname)
obj = {0: 0}
view.write_object('singleton', obj)
for iteration in range(10):
obj = {'iteration': iteration}
view.write_object(varname, obj, iteration=iteration)
for iteration in range(10):
obj = storage.get_object(envname,
modname,
varname,
iteration=iteration)
assert ('iteration' in obj)
assert (obj['iteration'] == iteration)
def test_write_array():
"""Test writing of a array.
"""
tmpfile = tempfile.NamedTemporaryFile()
storage = NetCDFStorage(tmpfile.name, mode='w')
view1 = NetCDFStorageView(storage, 'envname1', 'modname')
view2 = NetCDFStorageView(storage, 'envname2', 'modname')
from numpy.random import random
shape = (10, 3)
array = random(shape)
view1.write_array('singleton', array)
for iteration in range(10):
array = random(shape)
view1.write_array('varname', array, iteration=iteration)
view2.write_array('varname', array, iteration=iteration)
for iteration in range(10):
array = storage._ncfile['/envname1/modname/varname'][iteration]
assert array.shape == shape
array = storage._ncfile['/envname2/modname/varname'][iteration]
assert array.shape == shape
def __init__(self, target_samplers, storage=None, verbose=False):
"""
Initialize a multi-objective design sampler with the specified target sampler powers.
Parameters
----------
target_samplers : dict
target_samplers[sampler] is the exponent associated with SAMS sampler `sampler` in the multi-objective design.
storage : NetCDFStorage, optional, default=None
If specified, will use the storage layer to write trajectory data.
verbose : bool, optional, default=False
If true, will print verbose output
The target sampler weights for N samplers with specified exponents \alpha_n are given by
\pi_{nk} \propto \prod_{n=1}^N Z_{nk}^{alpha_n}
where \pi_{nk} is the target weight for sampler n state k,
and Z_{nk} is the relative partition function of sampler n among states k.
Examples
--------
Set up a mutation sampler to maximize implicit solvent hydration free energy.
>>> from perses.tests.testsystems import AlanineDipeptideTestSystem
>>> testsystem = AlanineDipeptideTestSystem()
>>> # Set up target samplers.
>>> target_samplers = { testsystem.sams_samplers['implicit'] : 1.0, testsystem.sams_samplers['vacuum'] : -1.0 }
>>> # Set up the design sampler.
>>> designer = MultiTargetDesign(target_samplers)
"""
# Store target samplers.
self.sampler_exponents = target_samplers
self.samplers = list(target_samplers.keys())
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize storage for target probabilities.
self.log_target_probabilities = dict()
self.verbose = verbose
self.iteration = 0
def test_write_quantity():
"""Test writing of a quantity.
"""
tmpfile = tempfile.NamedTemporaryFile()
storage = NetCDFStorage(tmpfile.name, mode='w')
view = NetCDFStorageView(storage, 'envname', 'modname')
view.write_quantity('singleton', 1.0)
for iteration in range(10):
view.write_quantity('varname', float(iteration), iteration=iteration)
for iteration in range(10):
assert (storage._ncfile['/envname/modname/varname'][iteration] == float(iteration))
def test_write_array():
"""Test writing of a array.
"""
tmpfile = tempfile.NamedTemporaryFile()
storage = NetCDFStorage(tmpfile.name, mode='w')
view1 = NetCDFStorageView(storage, 'envname1', 'modname')
view2 = NetCDFStorageView(storage, 'envname2', 'modname')
from numpy.random import random
shape = (10,3)
array = random(shape)
view1.write_array('singleton', array)
for iteration in range(10):
array = random(shape)
view1.write_array('varname', array, iteration=iteration)
view2.write_array('varname', array, iteration=iteration)
for iteration in range(10):
array = storage._ncfile['/envname1/modname/varname'][iteration]
assert array.shape == shape
array = storage._ncfile['/envname2/modname/varname'][iteration]
assert array.shape == shape
def test_write_object():
"""Test writing of a object.
"""
tmpfile = tempfile.NamedTemporaryFile()
storage = NetCDFStorage(tmpfile.name, mode='w')
view = NetCDFStorageView(storage, 'envname', 'modname')
obj = { 0 : 0 }
view.write_object('singleton', obj)
for iteration in range(10):
obj = { 'iteration' : iteration }
view.write_object('varname', obj, iteration=iteration)
for iteration in range(10):
obj = storage.get_object('/envname/modname/varname', iteration=iteration)
assert ('iteration' in obj)
assert (obj['iteration'] == iteration)
class ExpandedEnsembleSampler(object):
"""
Method of expanded ensembles sampling engine.
The acceptance criteria is given in the reference document. Roughly, the proposal scheme is:
* Draw a proposed chemical state k', and calculate reverse proposal probability
* Conditioned on k' and the current positions x, generate new positions with the GeometryEngine
* With new positions, jump to a hybrid system at lambda=0
* Anneal from lambda=0 to lambda=1, accumulating work
* Jump from the hybrid system at lambda=1 to the k' system, and compute reverse GeometryEngine proposal
* Add weight of chemical states k and k' to acceptance probabilities
Properties
----------
sampler : MCMCSampler
The MCMC sampler used for updating positions.
proposal_engine : ProposalEngine
The ProposalEngine to use for proposing new sampler states and topologies.
system_generator : SystemGenerator
The SystemGenerator to use for creating System objects following proposals.
state : hashable object
The current sampler state. Can be any hashable object.
states : set of hashable object
All known states.
iteration : int
Iterations completed.
naccepted : int
Number of accepted thermodynamic/chemical state changes.
nrejected : int
Number of rejected thermodynamic/chemical state changes.
number_of_state_visits : dict of state_key
Cumulative counts of visited states.
verbose : bool
If True, verbose output is printed.
References
----------
[1] Lyubartsev AP, Martsinovski AA, Shevkunov SV, and Vorontsov-Velyaminov PN. New approach to Monte Carlo calculation of the free energy: Method of expanded ensembles. JCP 96:1776, 1992
http://dx.doi.org/10.1063/1.462133
Examples
--------
>>> # Create a test system
>>> test = testsystems.AlanineDipeptideVacuum()
>>> # Create a SystemGenerator and rebuild the System.
>>> from perses.rjmc.topology_proposal import SystemGenerator
>>> system_generator = SystemGenerator(['amber99sbildn.xml'], forcefield_kwargs={ 'nonbondedMethod' : app.NoCutoff, 'implicitSolvent' : None, 'constraints' : None })
>>> test.system = system_generator.build_system(test.topology)
>>> # Create a sampler state.
>>> sampler_state = SamplerState(system=test.system, positions=test.positions)
>>> # Create a thermodynamic state.
>>> thermodynamic_state = ThermodynamicState(system=test.system, temperature=298.0*unit.kelvin)
>>> # Create an MCMC sampler
>>> mcmc_sampler = MCMCSampler(thermodynamic_state, sampler_state)
>>> # Turn off verbosity
>>> mcmc_sampler.verbose = False
>>> # Create an Expanded Ensemble sampler
>>> from perses.rjmc.topology_proposal import PointMutationEngine
>>> from perses.rjmc.geometry import FFAllAngleGeometryEngine
>>> geometry_engine = FFAllAngleGeometryEngine(metadata={})
>>> allowed_mutations = [[('2','ALA')],[('2','VAL'),('2','LEU')]]
>>> proposal_engine = PointMutationEngine(test.topology, system_generator, max_point_mutants=1, chain_id='1', proposal_metadata=None, allowed_mutations=allowed_mutations)
>>> exen_sampler = ExpandedEnsembleSampler(mcmc_sampler, test.topology, 'ACE-ALA-NME', proposal_engine, geometry_engine)
>>> # Run the sampler
>>> exen_sampler.run()
"""
def __init__(self, sampler, topology, state_key, proposal_engine, geometry_engine, log_weights=None, options=None, platform=None, envname=None, storage=None, ncmc_write_interval=1):
"""
Create an expanded ensemble sampler.
p(x,k) \propto \exp[-u_k(x) + g_k]
where g_k is the log weight.
Parameters
----------
sampler : MCMCSampler
MCMCSampler initialized with current SamplerState
topology : simtk.openmm.app.Topology
Current topology
state : hashable object
Current chemical state
proposal_engine : ProposalEngine
ProposalEngine to use for proposing new chemical states
geometry_engine : GeometryEngine
GeometryEngine to use for dimension matching
log_weights : dict of object : float
Log weights to use for expanded ensemble biases.
options : dict, optional, default=dict()
Options for initializing switching scheme, such as 'timestep', 'nsteps', 'functions' for NCMC
platform : simtk.openmm.Platform, optional, default=None
Platform to use for NCMC switching. If `None`, default (fastest) platform is used.
storage : NetCDFStorageView, optional, default=None
If specified, use this storage layer.
ncmc_write_interval : int, default 1
How frequently to write out NCMC protocol steps.
"""
# Keep copies of initializing arguments.
# TODO: Make deep copies?
self.sampler = sampler
self._pressure = sampler.thermodynamic_state.pressure
self._temperature = sampler.thermodynamic_state.temperature
self.topology = md.Topology.from_openmm(topology)
self.state_key = state_key
self.proposal_engine = proposal_engine
self.log_weights = log_weights
if self.log_weights is None: self.log_weights = dict()
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize
self.iteration = 0
option_names = ['timestep', 'nsteps', 'functions', 'nsteps_mcmc', 'splitting']
if options is None:
options = dict()
for option_name in option_names:
if option_name not in options:
options[option_name] = None
if options['splitting']:
self._ncmc_splitting = options['splitting']
else:
self._ncmc_splitting = "V R O H R V"
if options['nsteps']:
self._switching_nsteps = options['nsteps']
self.ncmc_engine = NCMCEngine(temperature=self.sampler.thermodynamic_state.temperature,
timestep=options['timestep'], nsteps=options['nsteps'],
functions=options['functions'], integrator_splitting=self._ncmc_splitting,
platform=platform, storage=self.storage,
write_ncmc_interval=ncmc_write_interval)
else:
self._switching_nsteps = 0
if options['nsteps_mcmc']:
self._n_iterations_per_update = options['nsteps_mcmc']
else:
self._n_iterations_per_update = 100
self.geometry_engine = geometry_engine
self.naccepted = 0
self.nrejected = 0
self.number_of_state_visits = dict()
self.verbose = False
self.pdbfile = None # if not None, write PDB file
self.geometry_pdbfile = None # if not None, write PDB file of geometry proposals
self.accept_everything = False # if True, will accept anything that doesn't lead to NaNs
self.logPs = list()
self.sampler.minimize(max_iterations=40)
@property
def state_keys(self):
return self.log_weights.keys()
def get_log_weight(self, state_key):
"""
Get the log weight of the specified state.
Parameters
----------
state_key : hashable object
The state key (e.g. chemical state key) to look up.
Returns
-------
log_weight : float
The log weight of the provided state key.
Note
----
This adds the key to the self.log_weights dict.
"""
if state_key not in self.log_weights:
self.log_weights[state_key] = 0.0
return self.log_weights[state_key]
def _system_to_thermodynamic_state(self, system):
"""
Given an OpenMM system object, create a corresponding ThermodynamicState that has the same
temperature and pressure as the current thermodynamic state.
Arguments
---------
system : openmm.System
The OpenMM system for which to create the thermodynamic state
Returns
-------
new_thermodynamic_state : openmmtools.states.ThermodynamicState
The thermodynamic state object representing the given system
"""
return ThermodynamicState(system, temperature=self._temperature, pressure=self._pressure)
def _geometry_forward(self, topology_proposal, old_sampler_state):
"""
Run geometry engine to propose new positions and compute logP
Parameters
----------
topology_proposal : TopologyProposal
Contains old/new Topology and System objects and atom mappings.
old_sampler_state : openmmtools.states.SamplerState
Configurational properties of the old system atoms.
Returns
-------
new_sampler_state : openmmtools.states.SamplerState
Configurational properties of new atoms proposed by geometry engine calculation.
geometry_logp_propose : float
The log probability of the forward-only proposal
"""
if self.verbose: print("Geometry engine proposal...")
# Generate coordinates for new atoms and compute probability ratio of old and new probabilities.
initial_time = time.time()
new_positions, geometry_logp_propose = self.geometry_engine.propose(topology_proposal, old_sampler_state.positions, self.sampler.thermodynamic_state.beta)
if self.verbose: print('proposal took %.3f s' % (time.time() - initial_time))
if self.geometry_pdbfile is not None:
print("Writing proposed geometry...")
from simtk.openmm.app import PDBFile
PDBFile.writeFile(topology_proposal.new_topology, new_positions, file=self.geometry_pdbfile)
self.geometry_pdbfile.flush()
new_sampler_state = SamplerState(new_positions, box_vectors=old_sampler_state.box_vectors)
return new_sampler_state, geometry_logp_propose
def _geometry_reverse(self, topology_proposal, new_sampler_state, old_sampler_state):
"""
Run geometry engine reverse calculation to determine logP
of proposing the old positions based on the new positions
Parameters
----------
topology_proposal : TopologyProposal
Contains old/new Topology and System objects and atom mappings.
new_sampler_state : openmmtools.states.SamplerState
Configurational properties of the new atoms.
old_sampler_state : openmmtools.states.SamplerState
Configurational properties of the old atoms.
Returns
-------
geometry_logp_reverse : float
The log probability of the proposal for the given transformation
"""
if self.verbose: print("Geometry engine logP_reverse calculation...")
initial_time = time.time()
geometry_logp_reverse = self.geometry_engine.logp_reverse(topology_proposal, new_sampler_state.positions, old_sampler_state.positions, self.sampler.thermodynamic_state.beta)
if self.verbose: print('calculation took %.3f s' % (time.time() - initial_time))
return geometry_logp_reverse
def _ncmc_hybrid(self, topology_proposal, old_sampler_state, new_sampler_state):
"""
Run a hybrid NCMC protocol from lambda = 0 to lambda = 1
Parameters
----------
topology_proposal : TopologyProposal
Contains old/new Topology and System objects and atom mappings.
old_sampler_State : openmmtools.states.SamplerState
SamplerState of old system at the beginning of NCMCSwitching
new_sampler_state : openmmtools.states.SamplerState
SamplerState of new system at the beginning of NCMCSwitching
Returns
-------
old_final_sampler_state : openmmtools.states.SamplerState
SamplerState of old system at the end of switching
new_final_sampler_state : openmmtools.states.SamplerState
SamplerState of new system at the end of switching
logP_work : float
The NCMC work contribution to the log acceptance probability (Eq. 44)
logP_energy : float
The contribution of switching to and from the hybrid system to the acceptance probability (Eq. 45)
"""
if self.verbose: print("Performing NCMC switching")
initial_time = time.time()
[ncmc_old_sampler_state, ncmc_new_sampler_state, logP_work, logP_initial_hybrid, logP_final_hybrid] = self.ncmc_engine.integrate(topology_proposal, old_sampler_state, new_sampler_state, iteration=self.iteration)
if self.verbose: print('NCMC took %.3f s' % (time.time() - initial_time))
# Check that positions are not NaN
if new_sampler_state.has_nan():
raise Exception("Positions are NaN after NCMC insert with %d steps" % self._switching_nsteps)
return ncmc_old_sampler_state, ncmc_new_sampler_state, logP_work, logP_initial_hybrid, logP_final_hybrid
def _geometry_ncmc_geometry(self, topology_proposal, sampler_state, old_log_weight, new_log_weight):
"""
Use a hybrid NCMC protocol to switch from the old system to new system
Will calculate new positions for the new system first, then give both
sets of positions to the hybrid NCMC integrator, and finally use the
final positions of the old and new systems to calculate the reverse
geometry probability
Parameters
----------
topology_proposal : TopologyProposal
Contains old/new Topology and System objects and atom mappings.
sampler_state : openmmtools.states.SamplerState
Configurational properties of old atoms at the beginning of the NCMC switching.
old_log_weight : float
Chemical state weight from SAMSSampler
new_log_weight : float
Chemical state weight from SAMSSampler
Returns
-------
logP_accept : float
Log of acceptance probability of entire Expanded Ensemble switch (Eq. 25 or 46)
ncmc_new_sampler_state : openmmtools.states.SamplerState
Configurational properties of new atoms at the end of the NCMC switching.
"""
if self.verbose: print("Updating chemical state with geometry-ncmc-geometry scheme...")
from perses.tests.utils import compute_potential
logP_chemical_proposal = topology_proposal.logp_proposal
old_thermodynamic_state = self.sampler.thermodynamic_state
new_thermodynamic_state = self._system_to_thermodynamic_state(topology_proposal.new_system)
initial_reduced_potential = feptasks.compute_reduced_potential(old_thermodynamic_state, sampler_state)
logP_initial_nonalchemical = - initial_reduced_potential
new_geometry_sampler_state, logP_geometry_forward = self._geometry_forward(topology_proposal, sampler_state)
#if we aren't doing any switching, then skip running the NCMC engine at all.
if self._switching_nsteps == 0:
ncmc_old_sampler_state = sampler_state
ncmc_new_sampler_state = new_geometry_sampler_state
logP_work = 0.0
logP_initial_hybrid = 0.0
logP_final_hybrid = 0.0
else:
ncmc_old_sampler_state, ncmc_new_sampler_state, logP_work, logP_initial_hybrid, logP_final_hybrid = self._ncmc_hybrid(topology_proposal, sampler_state, new_geometry_sampler_state)
if logP_work > -np.inf and logP_initial_hybrid > -np.inf and logP_final_hybrid > -np.inf:
logP_geometry_reverse = self._geometry_reverse(topology_proposal, ncmc_new_sampler_state, ncmc_old_sampler_state)
logP_to_hybrid = logP_initial_hybrid - logP_initial_nonalchemical
final_reduced_potential = feptasks.compute_reduced_potential(new_thermodynamic_state, ncmc_new_sampler_state)
logP_final_nonalchemical = -final_reduced_potential
logP_from_hybrid = logP_final_nonalchemical - logP_final_hybrid
logP_sams_weight = new_log_weight - old_log_weight
# Compute total log acceptance probability according to Eq. 46
logP_accept = logP_to_hybrid - logP_geometry_forward + logP_work + logP_from_hybrid + logP_geometry_reverse + logP_sams_weight
else:
logP_geometry_reverse = 0.0
logP_final = 0.0
logP_to_hybrid = 0.0
logP_from_hybrid = 0.0
logP_sams_weight = new_log_weight - old_log_weight
logP_accept = logP_to_hybrid - logP_geometry_forward + logP_work + logP_from_hybrid + logP_geometry_reverse + logP_sams_weight
#TODO: mark failed proposals as unproposable
if self.verbose:
print("logP_accept = %+10.4e [logP_to_hybrid = %+10.4e, logP_chemical_proposal = %10.4e, logP_reverse = %+10.4e, -logP_forward = %+10.4e, logP_work = %+10.4e, logP_from_hybrid = %+10.4e, logP_sams_weight = %+10.4e]"
% (logP_accept, logP_to_hybrid, logP_chemical_proposal, logP_geometry_reverse, -logP_geometry_forward, logP_work, logP_from_hybrid, logP_sams_weight))
# Write to storage.
if self.storage:
self.storage.write_quantity('logP_accept', logP_accept, iteration=self.iteration)
# Write components to storage
self.storage.write_quantity('logP_ncmc_work', logP_work, iteration=self.iteration)
self.storage.write_quantity('logP_from_hybrid', logP_from_hybrid, iteration=self.iteration)
self.storage.write_quantity('logP_to_hybrid', logP_to_hybrid, iteration=self.iteration)
self.storage.write_quantity('logP_chemical_proposal', logP_chemical_proposal, iteration=self.iteration)
self.storage.write_quantity('logP_reverse', logP_geometry_reverse, iteration=self.iteration)
self.storage.write_quantity('logP_forward', logP_geometry_forward, iteration=self.iteration)
self.storage.write_quantity('logP_sams_weight', logP_sams_weight, iteration=self.iteration)
# Write some aggregate statistics to storage to make contributions to acceptance probability easier to analyze
self.storage.write_quantity('logP_groups_chemical', logP_chemical_proposal, iteration=self.iteration)
self.storage.write_quantity('logP_groups_geometry', logP_geometry_reverse - logP_geometry_forward, iteration=self.iteration)
return logP_accept, ncmc_new_sampler_state
def update_positions(self, n_iterations=1):
"""
Sample new positions.
"""
self.sampler.run(n_iterations=n_iterations)
def update_state(self):
"""
Sample the thermodynamic state.
"""
initial_time = time.time()
# Propose new chemical state.
if self.verbose: print("Proposing new topology...")
[system, topology, positions] = [self.sampler.thermodynamic_state.get_system(remove_thermostat=True), self.topology, self.sampler.sampler_state.positions]
omm_topology = topology.to_openmm() #convert to OpenMM topology for proposal engine
omm_topology.setPeriodicBoxVectors(self.sampler.sampler_state.box_vectors) #set the box vectors because in OpenMM topology has these...
topology_proposal = self.proposal_engine.propose(system, omm_topology)
if self.verbose: print("Proposed transformation: %s => %s" % (topology_proposal.old_chemical_state_key, topology_proposal.new_chemical_state_key))
# Determine state keys
old_state_key = self.state_key
new_state_key = topology_proposal.new_chemical_state_key
# Determine log weight
old_log_weight = self.get_log_weight(old_state_key)
new_log_weight = self.get_log_weight(new_state_key)
logp_accept, ncmc_new_sampler_state = self._geometry_ncmc_geometry(topology_proposal, self.sampler.sampler_state, old_log_weight, new_log_weight)
# Accept or reject.
if np.isnan(logp_accept):
accept = False
print('logp_accept = NaN')
else:
accept = ((logp_accept>=0.0) or (np.random.uniform() < np.exp(logp_accept)))
if self.accept_everything:
print('accept_everything option is turned on; accepting')
accept = True
if accept:
self.sampler.thermodynamic_state.set_system(topology_proposal.new_system, fix_state=True)
self.sampler.sampler_state.system = topology_proposal.new_system
self.topology = md.Topology.from_openmm(topology_proposal.new_topology)
self.sampler.sampler_state = ncmc_new_sampler_state
self.sampler.topology = self.topology
self.state_key = topology_proposal.new_chemical_state_key
self.naccepted += 1
if self.verbose: print(" accepted")
else:
self.nrejected += 1
if self.verbose: print(" rejected")
if self.storage:
self.storage.write_configuration('positions', self.sampler.sampler_state.positions, self.topology, iteration=self.iteration)
self.storage.write_object('state_key', self.state_key, iteration=self.iteration)
self.storage.write_object('proposed_state_key', topology_proposal.new_chemical_state_key, iteration=self.iteration)
self.storage.write_quantity('naccepted', self.naccepted, iteration=self.iteration)
self.storage.write_quantity('nrejected', self.nrejected, iteration=self.iteration)
self.storage.write_quantity('logp_accept', logp_accept, iteration=self.iteration)
self.storage.write_quantity('logp_topology_proposal', topology_proposal.logp_proposal, iteration=self.iteration)
# Update statistics.
self.update_statistics()
def update(self):
"""
Update the sampler with one step of sampling.
"""
if self.verbose:
print("-" * 80)
print("Expanded Ensemble sampler iteration %8d" % self.iteration)
self.update_positions(n_iterations=self._n_iterations_per_update)
self.update_state()
self.iteration += 1
if self.verbose:
print("-" * 80)
if self.pdbfile is not None:
print("Writing frame...")
from simtk.openmm.app import PDBFile
PDBFile.writeModel(self.topology.to_openmm(), self.sampler.sampler_state.positions, self.pdbfile, self.iteration)
self.pdbfile.flush()
if self.storage:
self.storage.sync()
def run(self, niterations=1):
"""
Run the sampler for the specified number of iterations
Parameters
----------
niterations : int, optional, default=1
Number of iterations to run the sampler for.
"""
for iteration in range(niterations):
self.update()
def update_statistics(self):
"""
Update sampler statistics.
"""
if self.state_key not in self.number_of_state_visits:
self.number_of_state_visits[self.state_key] = 0
self.number_of_state_visits[self.state_key] += 1
def __init__(self, sampler, logZ=None, log_target_probabilities=None, update_method='two-stage', storage=None, second_stage_start=1000):
"""
Create a SAMS Sampler.
Parameters
----------
sampler : ExpandedEnsembleSampler
The expanded ensemble sampler used to sample both configurations and discrete thermodynamic states.
logZ : dict of key : float, optional, default=None
If specified, the log partition functions for each state will be initialized to the specified dictionary.
log_target_probabilities : dict of key : float, optional, default=None
If specified, unnormalized target probabilities; default is all 0.
update_method : str, optional, default='default'
SAMS update algorithm
storage : NetCDFStorageView, optional, default=None
second_state_start : int, optional, default None
At what iteration number to switch to the optimal gain decay
"""
from scipy.misc import logsumexp
from perses.tests.utils import createOEMolFromSMILES
# Keep copies of initializing arguments.
# TODO: Make deep copies?
self.sampler = sampler
self.chemical_states = None
self._reference_state = None
try:
self.chemical_states = self.sampler.proposal_engine.chemical_state_list
except NotImplementedError:
logger.warn("The proposal engine has not properly implemented the chemical state property; SAMS will add states on the fly.")
if self.chemical_states:
# Select a reference state that will always be subtracted (ensure that dict ordering does not change)
self._reference_state = self.chemical_states[0]
# Initialize the logZ dictionary with scores based on the number of atoms
# This is not the negative because the weights are set to the negative of the initial weights
self.logZ = {chemical_state: self._num_dof_compensation(chemical_state) for chemical_state in self.chemical_states}
#Initialize log target probabilities with log(1/n_states)
self.log_target_probabilities = {chemical_state : np.log(len(self.chemical_states)) for chemical_state in self.chemical_states}
#If initial weights are specified, override any weight with what is provided
#However, if the chemical state is not in the reachable chemical state list,throw an exception
if logZ is not None:
for (chemical_state, logZ_value) in logZ:
if chemical_state not in self.chemical_states:
raise ValueError("Provided a logZ initial value for an un-proposable chemical state")
self.logZ[chemical_state] = logZ_value
if log_target_probabilities is not None:
for (chemical_state, log_target_probability) in log_target_probabilities:
if chemical_state not in self.chemical_states:
raise ValueError("Provided a log target probability for an un-proposable chemical state.")
self.log_target_probabilities[chemical_state] = log_target_probability
#normalize target probabilities
#this is likely not necessary, but it is copying the algorithm in Ref 1
log_sum_target_probabilities = logsumexp((list(self.log_target_probabilities.values())))
self.log_target_probabilities = {chemical_state : log_target_probability - log_sum_target_probabilities for chemical_state, log_target_probability in self.log_target_probabilities}
else:
self.logZ = dict()
self.log_target_probabilities = dict()
self.update_method = update_method
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize.
self.iteration = 0
self.verbose = False
self.sampler.log_weights = {state_key: - self.logZ[state_key] for state_key in self.logZ.keys()}
self.second_stage_start = 0
if second_stage_start is not None:
self.second_stage_start = second_stage_start
class ProtonationStateSampler(object):
"""
Protonation state sampler with given fixed target probabilities for ligand in solvent.
Parameters
----------
samplers : list of SAMSSampler
The SAMS samplers whose relative partition functions go into the design objective computation.
sampler_exponents : dict of SAMSSampler : float
samplers.keys() are the samplers, and samplers[key]
log_target_probabilities : dict of hashable object : float
log_target_probabilities[key] is the computed log objective function (target probability) for chemical state `key`
verbose : bool
If True, verbose output is printed.
"""
def __init__(self, complex_sampler, solvent_sampler, log_state_penalties, storage=None, verbose=False):
"""
Initialize a protonation state sampler with fixed target probabilities for ligand in solvent.
Parameters
----------
complex_sampler : ExpandedEnsembleSampler
Ligand in complex sampler
solvent_sampler : SAMSSampler
Ligand in solution sampler
log_state_penalties : dict
log_state_penalties[smiles] is the log state free energy (in kT) for ligand state 'smiles'
storage : NetCDFStorage, optional, default=None
If specified, will use the storage layer to write trajectory data.
verbose : bool, optional, default=False
If true, will print verbose output
"""
# Store target samplers.
self.log_state_penalties = log_state_penalties
self.samplers = [complex_sampler, solvent_sampler]
self.complex_sampler = complex_sampler
self.solvent_sampler = solvent_sampler
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize storage for target probabilities.
self.log_target_probabilities = { key : - log_state_penalties[key] for key in log_state_penalties }
self.verbose = verbose
self.iteration = 0
@property
def state_keys(self):
return self.log_target_probabilities.keys()
def update_samplers(self):
"""
Update all samplers.
"""
for sampler in self.samplers:
sampler.update()
def update_target_probabilities(self):
"""
Update all target probabilities.
"""
# Update the complex sampler log weights using the solvent sampler log weights
for key in self.solvent_sampler.state_keys:
self.complex_sampler.log_weights[key] = self.solvent_sampler.sampler.log_weights[key]
if self.verbose:
print("log_weights = %s" % str(self.solvent_sampler.sampler.log_weights))
def update(self):
"""
Run one iteration of the sampler.
"""
if self.verbose:
print("*" * 80)
print("ProtonationStateSampler iteration %8d" % self.iteration)
self.update_samplers()
self.update_target_probabilities()
if self.storage: self.storage.sync()
self.iteration += 1
if self.verbose:
print("*" * 80)
def run(self, niterations=1):
"""
Run the protonation state sampler for the specified number of iterations.
Parameters
----------
niterations : int
The number of iterations to run the sampler for.
"""
# Update all samplers.
for iteration in range(niterations):
self.update()
class MultiTargetDesign(object):
"""
Multi-objective design using self-adjusted mixture sampling with additional recursion steps
that update target weights on the fly.
Parameters
----------
samplers : list of SAMSSampler
The SAMS samplers whose relative partition functions go into the design objective computation.
sampler_exponents : dict of SAMSSampler : float
samplers.keys() are the samplers, and samplers[key]
log_target_probabilities : dict of hashable object : float
log_target_probabilities[key] is the computed log objective function (target probability) for chemical state `key`
verbose : bool
If True, verbose output is printed.
"""
def __init__(self, target_samplers, storage=None, verbose=False):
"""
Initialize a multi-objective design sampler with the specified target sampler powers.
Parameters
----------
target_samplers : dict
target_samplers[sampler] is the exponent associated with SAMS sampler `sampler` in the multi-objective design.
storage : NetCDFStorage, optional, default=None
If specified, will use the storage layer to write trajectory data.
verbose : bool, optional, default=False
If true, will print verbose output
The target sampler weights for N samplers with specified exponents \alpha_n are given by
\pi_{nk} \propto \prod_{n=1}^N Z_{nk}^{alpha_n}
where \pi_{nk} is the target weight for sampler n state k,
and Z_{nk} is the relative partition function of sampler n among states k.
Examples
--------
Set up a mutation sampler to maximize implicit solvent hydration free energy.
>>> from perses.tests.testsystems import AlanineDipeptideTestSystem
>>> testsystem = AlanineDipeptideTestSystem()
>>> # Set up target samplers.
>>> target_samplers = { testsystem.sams_samplers['implicit'] : 1.0, testsystem.sams_samplers['vacuum'] : -1.0 }
>>> # Set up the design sampler.
>>> designer = MultiTargetDesign(target_samplers)
"""
# Store target samplers.
self.sampler_exponents = target_samplers
self.samplers = list(target_samplers.keys())
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize storage for target probabilities.
self.log_target_probabilities = dict()
self.verbose = verbose
self.iteration = 0
@property
def state_keys(self):
return self.log_target_probabilities.keys()
def update_samplers(self):
"""
Update all samplers.
"""
for sampler in self.samplers:
sampler.update()
def update_target_probabilities(self):
"""
Update all target probabilities.
"""
# Gather list of all keys.
state_keys = set()
for sampler in self.samplers:
for key in sampler.state_keys:
state_keys.add(key)
# Compute unnormalized log target probabilities.
log_target_probabilities = { key : 0.0 for key in state_keys }
for (sampler, log_weight) in self.sampler_exponents.items():
for key in sampler.state_keys:
log_target_probabilities[key] += log_weight * sampler.logZ[key]
# Normalize
log_sum = log_sum_exp(log_target_probabilities)
for key in log_target_probabilities:
log_target_probabilities[key] -= log_sum
# Store.
self.log_target_probabilities = log_target_probabilities
if self.verbose:
print("log_target_probabilities = %s" % str(self.log_target_probabilities))
if self.storage:
self.storage.write_object('log_target_probabilities', self.log_target_probabilities, iteration=self.iteration)
def update(self):
"""
Run one iteration of the sampler.
"""
if self.verbose:
print("*" * 80)
print("MultiTargetDesign sampler iteration %8d" % self.iteration)
self.update_samplers()
self.update_target_probabilities()
self.iteration += 1
if self.storage: self.storage.sync()
if self.verbose:
print("*" * 80)
def run(self, niterations=1):
"""
Run the multi-target design sampler for the specified number of iterations.
Parameters
----------
niterations : int
The number of iterations to run the sampler for.
"""
# Update all samplers.
for iteration in range(niterations):
self.update()
def __init__(self, temperature=default_temperature, functions=None, nsteps=default_nsteps,
steps_per_propagation=default_steps_per_propagation, timestep=default_timestep,
constraint_tolerance=None, platform=None, write_ncmc_interval=1, measure_shadow_work=False,
integrator_splitting='V R O H R V', storage=None, verbose=False, LRUCapacity=10, pressure=None, bond_softening_constant=1.0, angle_softening_constant=1.0):
"""
This is the base class for NCMC switching between two different systems.
Arguments
---------
temperature : simtk.unit.Quantity with units compatible with kelvin
The temperature at which switching is to be run
functions : dict of str:str, optional, default=default_functions
functions[parameter] is the function (parameterized by 't' which switched from 0 to 1) that
controls how alchemical context parameter 'parameter' is switched
nsteps : int, optional, default=1
The number of steps to use for switching.
steps_per_propagation : int, optional, default=1
The number of intermediate propagation steps taken at each switching step
timestep : simtk.unit.Quantity with units compatible with femtoseconds, optional, default=1*femtosecond
The timestep to use for integration of switching velocity Verlet steps.
constraint_tolerance : float, optional, default=None
If not None, this relative constraint tolerance is used for position and velocity constraints.
platform : simtk.openmm.Platform, optional, default=None
If specified, the platform to use for OpenMM simulations.
write_ncmc_interval : int, optional, default=None
If a positive integer is specified, a snapshot frame will be written to storage with the specified interval on NCMC switching.
'storage' must also be specified.
measure_shadow_work : bool, optional, default False
Whether to measure shadow work
integrator_splitting : str, optional, default='V R O H R V'
NCMC internal integrator splitting based on OpenMMTools Langevin splittings
storage : NetCDFStorageView, optional, default=None
If specified, write data using this class.
verbose : bool, optional, default=False
If True, print debug information.
LRUCapacity : int, default 10
Capacity of LRU cache for hybrid systems
pressure : float, default None
The pressure to use for the simulation. If None, no barostat
"""
# Handle some defaults.
if functions == None:
functions = python_hybrid_functions
if nsteps == None:
nsteps = default_nsteps
if timestep == None:
timestep = default_timestep
if temperature == None:
temperature = default_temperature
self._temperature = temperature
self._functions = copy.deepcopy(functions)
self._nsteps = nsteps
self._timestep = timestep
self._constraint_tolerance = constraint_tolerance
self._platform = platform
self._integrator_splitting = integrator_splitting
self._steps_per_propagation = steps_per_propagation
self._verbose = verbose
self._pressure = pressure
self._bond_softening_constant = bond_softening_constant
self._angle_softening_constant = angle_softening_constant
self._disable_barostat = False
self._hybrid_cache = LRUCache(capacity=LRUCapacity)
self._measure_shadow_work = measure_shadow_work
self._nattempted = 0
self._storage = None
if storage is not None:
self._storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
self._save_configuration = True
else:
self._save_configuration = False
if write_ncmc_interval is not None:
self._write_ncmc_interval = write_ncmc_interval
else:
self._write_ncmc_interval = 1
self._work_save_interval = write_ncmc_interval
class MultiTargetDesign(object):
"""
Multi-objective design using self-adjusted mixture sampling with additional recursion steps
that update target weights on the fly.
Parameters
----------
samplers : list of SAMSSampler
The SAMS samplers whose relative partition functions go into the design objective computation.
sampler_exponents : dict of SAMSSampler : float
samplers.keys() are the samplers, and samplers[key]
log_target_probabilities : dict of hashable object : float
log_target_probabilities[key] is the computed log objective function (target probability) for chemical state `key`
verbose : bool
If True, verbose output is printed.
"""
def __init__(self, target_samplers, storage=None, verbose=False):
"""
Initialize a multi-objective design sampler with the specified target sampler powers.
Parameters
----------
target_samplers : dict
target_samplers[sampler] is the exponent associated with SAMS sampler `sampler` in the multi-objective design.
storage : NetCDFStorage, optional, default=None
If specified, will use the storage layer to write trajectory data.
verbose : bool, optional, default=False
If true, will print verbose output
The target sampler weights for N samplers with specified exponents \alpha_n are given by
\pi_{nk} \propto \prod_{n=1}^N Z_{nk}^{alpha_n}
where \pi_{nk} is the target weight for sampler n state k,
and Z_{nk} is the relative partition function of sampler n among states k.
Examples
--------
Set up a mutation sampler to maximize implicit solvent hydration free energy.
>>> from perses.tests.testsystems import AlanineDipeptideTestSystem
>>> testsystem = AlanineDipeptideTestSystem()
>>> # Set up target samplers.
>>> target_samplers = { testsystem.sams_samplers['implicit'] : 1.0, testsystem.sams_samplers['vacuum'] : -1.0 }
>>> # Set up the design sampler.
>>> designer = MultiTargetDesign(target_samplers)
"""
# Store target samplers.
self.sampler_exponents = target_samplers
self.samplers = list(target_samplers.keys())
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize storage for target probabilities.
self.log_target_probabilities = dict()
self.verbose = verbose
self.iteration = 0
@property
def state_keys(self):
return self.log_target_probabilities.keys()
def update_samplers(self):
"""
Update all samplers.
"""
for sampler in self.samplers:
sampler.update()
def update_target_probabilities(self):
"""
Update all target probabilities.
"""
# Gather list of all keys.
state_keys = set()
for sampler in self.samplers:
for key in sampler.state_keys:
state_keys.add(key)
# Compute unnormalized log target probabilities.
log_target_probabilities = { key : 0.0 for key in state_keys }
for (sampler, log_weight) in self.sampler_exponents.items():
for key in sampler.state_keys:
log_target_probabilities[key] += log_weight * sampler.logZ[key]
# Normalize
log_sum = log_sum_exp(log_target_probabilities)
for key in log_target_probabilities:
log_target_probabilities[key] -= log_sum
# Store.
self.log_target_probabilities = log_target_probabilities
if self.verbose:
print("log_target_probabilities = %s" % str(self.log_target_probabilities))
if self.storage:
self.storage.write_object('log_target_probabilities', self.log_target_probabilities, iteration=self.iteration)
def update(self):
"""
Run one iteration of the sampler.
"""
if self.verbose:
print("*" * 80)
print("MultiTargetDesign sampler iteration %8d" % self.iteration)
self.update_samplers()
self.update_target_probabilities()
self.iteration += 1
if self.storage: self.storage.sync()
if self.verbose:
print("*" * 80)
def run(self, niterations=1):
"""
Run the multi-target design sampler for the specified number of iterations.
Parameters
----------
niterations : int
The number of iterations to run the sampler for.
"""
# Update all samplers.
for iteration in range(niterations):
self.update()
class SAMSSampler(object):
"""
Self-adjusted mixture sampling engine.
Properties
----------
state_keys : set of objects
The names of states sampled by the sampler.
logZ : dict() of keys : float
logZ[key] is the log partition function (up to an additive constant) estimate for chemical state `key`
update_method : str
Update method. One of ['default']
iteration : int
Iterations completed.
verbose : bool
If True, verbose debug output is printed.
References
----------
[1] Tan, Z. (2015) Optimally adjusted mixture sampling and locally weighted histogram analysis, Journal of Computational and Graphical Statistics, to appear. (Supplement)
http://www.stat.rutgers.edu/home/ztan/Publication/SAMS_redo4.pdf
Examples
--------
>>> # Create a test system
>>> test = testsystems.AlanineDipeptideVacuum()
>>> # Create a SystemGenerator and rebuild the System.
>>> from perses.rjmc.topology_proposal import SystemGenerator
>>> system_generator = SystemGenerator(['amber99sbildn.xml'], forcefield_kwargs={'implicitSolvent' : None, 'constraints' : None }, nonperiodic_forcefield_kwargs={'nonbondedMethod' : app.NoCutoff})
>>> test.system = system_generator.build_system(test.topology)
>>> # Create a sampler state.
>>> sampler_state = SamplerState(system=test.system, positions=test.positions)
>>> # Create a thermodynamic state.
>>> thermodynamic_state = ThermodynamicState(system=test.system, temperature=298.0*unit.kelvin)
>>> # Create an MCMC sampler
>>> mcmc_sampler = MCMCSampler(thermodynamic_state, sampler_state)
>>> # Turn off verbosity
>>> mcmc_sampler.verbose = False
>>> from perses.rjmc.geometry import FFAllAngleGeometryEngine
>>> geometry_engine = FFAllAngleGeometryEngine(metadata={})
>>> # Create an Expanded Ensemble sampler
>>> from perses.rjmc.topology_proposal import PointMutationEngine
>>> allowed_mutations = [[('2','ALA')],[('2','VAL'),('2','LEU')]]
>>> proposal_engine = PointMutationEngine(test.topology, system_generator, max_point_mutants=1, chain_id='1', proposal_metadata=None, allowed_mutations=allowed_mutations)
>>> exen_sampler = ExpandedEnsembleSampler(mcmc_sampler, test.topology, 'ACE-ALA-NME', proposal_engine, geometry_engine)
>>> # Create a SAMS sampler
>>> sams_sampler = SAMSSampler(exen_sampler)
>>> # Run the sampler
>>> sams_sampler.run() # doctest: +ELLIPSIS
...
"""
def __init__(self, sampler, logZ=None, log_target_probabilities=None, update_method='two-stage', storage=None, second_stage_start=1000):
"""
Create a SAMS Sampler.
Parameters
----------
sampler : ExpandedEnsembleSampler
The expanded ensemble sampler used to sample both configurations and discrete thermodynamic states.
logZ : dict of key : float, optional, default=None
If specified, the log partition functions for each state will be initialized to the specified dictionary.
log_target_probabilities : dict of key : float, optional, default=None
If specified, unnormalized target probabilities; default is all 0.
update_method : str, optional, default='default'
SAMS update algorithm
storage : NetCDFStorageView, optional, default=None
second_state_start : int, optional, default None
At what iteration number to switch to the optimal gain decay
"""
from scipy.special import logsumexp
from perses.utils.openeye import smiles_to_oemol
# Keep copies of initializing arguments.
# TODO: Make deep copies?
self.sampler = sampler
self.chemical_states = None
self._reference_state = None
try:
self.chemical_states = self.sampler.proposal_engine.chemical_state_list
except NotImplementedError:
_logger.warning("The proposal engine has not properly implemented the chemical state property; SAMS will add states on the fly.")
if self.chemical_states:
# Select a reference state that will always be subtracted (ensure that dict ordering does not change)
self._reference_state = self.chemical_states[0]
# Initialize the logZ dictionary with scores based on the number of atoms
# This is not the negative because the weights are set to the negative of the initial weights
self.logZ = {chemical_state: self._num_dof_compensation(chemical_state) for chemical_state in self.chemical_states}
#Initialize log target probabilities with log(1/n_states)
self.log_target_probabilities = {chemical_state : np.log(len(self.chemical_states)) for chemical_state in self.chemical_states}
#If initial weights are specified, override any weight with what is provided
#However, if the chemical state is not in the reachable chemical state list,throw an exception
if logZ is not None:
for (chemical_state, logZ_value) in logZ:
if chemical_state not in self.chemical_states:
raise ValueError("Provided a logZ initial value for an un-proposable chemical state")
self.logZ[chemical_state] = logZ_value
if log_target_probabilities is not None:
for (chemical_state, log_target_probability) in log_target_probabilities:
if chemical_state not in self.chemical_states:
raise ValueError("Provided a log target probability for an un-proposable chemical state.")
self.log_target_probabilities[chemical_state] = log_target_probability
#normalize target probabilities
#this is likely not necessary, but it is copying the algorithm in Ref 1
log_sum_target_probabilities = logsumexp((list(self.log_target_probabilities.values())))
self.log_target_probabilities = {chemical_state : log_target_probability - log_sum_target_probabilities for chemical_state, log_target_probability in self.log_target_probabilities}
else:
self.logZ = dict()
self.log_target_probabilities = dict()
self.update_method = update_method
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize.
self.iteration = 0
self.verbose = False
self.sampler.log_weights = {state_key: - self.logZ[state_key] for state_key in self.logZ.keys()}
self.second_stage_start = 0
if second_stage_start is not None:
self.second_stage_start = second_stage_start
@property
def state_keys(self):
return self.logZ.keys()
def _num_dof_compensation(self, smiles):
"""
Compute an approximate compensating factor for a chemical state based on the number of degrees of freedom that it has.
The formula is:
(num_heavy*heavy_factor) + (num_hydrogen*h_factor) where
heavy_factor = 4.5 and
light_factor = 3.8
Parameters
----------
smiles : str
The SMILES string of the molecule
Returns
-------
correction_factor : float
"""
mol = smiles_to_oemol(smiles)
num_heavy = 0
num_light = 0
heavy_factor = 4.5
light_factor = 3.8
for atom in mol.GetAtoms():
if atom.GetAtomicNum() == 1:
num_light += 1
else:
num_heavy += 1
correction_factor = num_heavy*heavy_factor + num_light*light_factor
return correction_factor
def update_sampler(self):
"""
Update the underlying expanded ensembles sampler.
"""
self.sampler.update()
def update_logZ_estimates(self):
"""
Update the logZ estimates according to self.update_method.
"""
state_key = self.sampler.state_key
# Add state key to dictionaries if we haven't visited this state before.
if state_key not in self.logZ:
_logger.warning("A new state key is being added to the logZ; note that this makes the resultant algorithm different from SAMS")
self.logZ[state_key] = 0.0
if state_key not in self.log_target_probabilities:
_logger.warning("A new state key is being added to the target probabilities; note that this makes the resultant algorithm different from SAMS")
self.log_target_probabilities[state_key] = 0.0
# Update estimates of logZ.
if self.update_method == 'one-stage':
# Based on Eq. 9 of Ref. [1]
gamma = 1.0 / float(self.iteration+1)
elif self.update_method == 'two-stage':
# Keep gamma large until second stage is activated.
if self.iteration < self.second_stage_start:
# First stage.
gamma = 1.0
# TODO: Determine when to switch to second stage
else:
# Second stage.
gamma = 1.0 / float(self.iteration - self.second_stage_start + 1)
else:
raise Exception("SAMS update method '%s' unknown." % self.update_method)
#get the (t-1/2) update from equation 9 in ref 1
self.logZ[state_key] += gamma / np.exp(self.log_target_probabilities[state_key])
if self._reference_state:
#the second step of the (t-1/2 update), subtracting the reference state from everything else.
#we can only do this for cases where all states have been enumerated
self.logZ = {state_key : logZ_estimate - self.logZ[self._reference_state] for state_key, logZ_estimate in self.logZ.items()}
# Update log weights for sampler.
self.sampler.log_weights = { state_key : - self.logZ[state_key] for state_key in self.logZ.keys()}
if self.storage:
self.storage.write_object('logZ', self.logZ, iteration=self.iteration)
self.storage.write_object('log_weights', self.sampler.log_weights, iteration=self.iteration)
def update(self):
"""
Update the sampler with one step of sampling.
"""
if self.verbose:
print("=" * 80)
print("SAMS sampler iteration %5d" % self.iteration)
self.update_sampler()
self.update_logZ_estimates()
if self.storage: self.storage.sync()
self.iteration += 1
if self.verbose:
print("=" * 80)
def run(self, niterations=1):
"""
Run the sampler for the specified number of iterations
Parameters
----------
niterations : int, optional, default=1
Number of iterations to run the sampler for.
"""
for iteration in range(niterations):
self.update()
def __init__(self, **kwargs):
super(Selectivity, self).__init__(**kwargs)
solvents = ['explicit'] # DEBUG
components = ['src-imatinib', 'abl-imatinib'] # TODO: Add 'ATP:kinase' complex to enable resistance design
padding = 9.0*unit.angstrom
explicit_solvent_model = 'tip3p'
setup_path = 'data/abl-src'
thermodynamic_states = dict()
temperature = 300*unit.kelvin
pressure = 1.0*unit.atmospheres
geometry_engine = geometry.FFAllAngleGeometryEngine()
# Construct list of all environments
environments = list()
for solvent in solvents:
for component in components:
environment = solvent + '-' + component
environments.append(environment)
# Read SMILES from CSV file of clinical kinase inhibitors.
from pkg_resources import resource_filename
smiles_filename = resource_filename('perses', 'data/clinical-kinase-inhibitors.csv')
import csv
molecules = list()
with open(smiles_filename, 'r') as csvfile:
csvreader = csv.reader(csvfile, delimiter=',', quotechar='"')
for row in csvreader:
name = row[0]
smiles = row[1]
molecules.append(smiles)
# Add current molecule
molecules.append('Cc1ccc(cc1Nc2nccc(n2)c3cccnc3)NC(=O)c4ccc(cc4)C[NH+]5CCN(CC5)C')
self.molecules = molecules
# Expand molecules without explicit stereochemistry and make canonical isomeric SMILES.
molecules = sanitizeSMILES(self.molecules)
# Create a system generator for desired forcefields
from perses.rjmc.topology_proposal import SystemGenerator
from pkg_resources import resource_filename
gaff_xml_filename = resource_filename('perses', 'data/gaff.xml')
system_generators = dict()
system_generators['explicit'] = SystemGenerator([gaff_xml_filename, 'amber99sbildn.xml', 'tip3p.xml'],
forcefield_kwargs={ 'nonbondedMethod' : app.CutoffPeriodic, 'nonbondedCutoff' : 9.0 * unit.angstrom, 'implicitSolvent' : None, 'constraints' : None },
use_antechamber=True)
# NOTE implicit solvent not supported by this SystemGenerator
# system_generators['implicit'] = SystemGenerator([gaff_xml_filename, 'amber99sbildn.xml', 'amber99_obc.xml'],
# forcefield_kwargs={ 'nonbondedMethod' : app.NoCutoff, 'implicitSolvent' : app.OBC2, 'constraints' : None },
# use_antechamber=True)
system_generators['vacuum'] = SystemGenerator([gaff_xml_filename, 'amber99sbildn.xml'],
forcefield_kwargs={ 'nonbondedMethod' : app.NoCutoff, 'implicitSolvent' : None, 'constraints' : None },
use_antechamber=True)
# Copy system generators for all environments
for solvent in solvents:
for component in components:
environment = solvent + '-' + component
system_generators[environment] = system_generators[solvent]
# Load topologies and positions for all components
from simtk.openmm.app import PDBFile, Modeller
topologies = dict()
positions = dict()
for component in components:
pdb_filename = resource_filename('perses', os.path.join(setup_path, '%s.pdb' % component))
print(pdb_filename)
pdbfile = PDBFile(pdb_filename)
topologies[component] = pdbfile.topology
positions[component] = pdbfile.positions
# Construct positions and topologies for all solvent environments
for solvent in solvents:
for component in components:
environment = solvent + '-' + component
if solvent == 'explicit':
# Create MODELLER object.
modeller = app.Modeller(topologies[component], positions[component])
modeller.addSolvent(system_generators[solvent].getForceField(), model='tip3p', padding=9.0*unit.angstrom)
topologies[environment] = modeller.getTopology()
positions[environment] = modeller.getPositions()
else:
environment = solvent + '-' + component
topologies[environment] = topologies[component]
positions[environment] = positions[component]
# Set up the proposal engines.
from perses.rjmc.topology_proposal import SmallMoleculeSetProposalEngine
proposal_metadata = { }
proposal_engines = dict()
for environment in environments:
storage = None
if self.storage:
storage = NetCDFStorageView(self.storage, envname=environment)
proposal_engines[environment] = SmallMoleculeSetProposalEngine(molecules, system_generators[environment], residue_name='MOL', storage=storage)
# Generate systems
systems = dict()
for environment in environments:
systems[environment] = system_generators[environment].build_system(topologies[environment])
# Define thermodynamic state of interest.
from perses.samplers.thermodynamics import ThermodynamicState
thermodynamic_states = dict()
temperature = 300*unit.kelvin
pressure = 1.0*unit.atmospheres
for component in components:
for solvent in solvents:
environment = solvent + '-' + component
if solvent == 'explicit':
thermodynamic_states[environment] = ThermodynamicState(system=systems[environment], temperature=temperature, pressure=pressure)
else:
thermodynamic_states[environment] = ThermodynamicState(system=systems[environment], temperature=temperature)
# Create SAMS samplers
from perses.samplers.samplers import SamplerState, MCMCSampler, ExpandedEnsembleSampler, SAMSSampler
mcmc_samplers = dict()
exen_samplers = dict()
sams_samplers = dict()
for solvent in solvents:
for component in components:
environment = solvent + '-' + component
chemical_state_key = proposal_engines[environment].compute_state_key(topologies[environment])
storage = None
if self.storage:
storage = NetCDFStorageView(self.storage, envname=environment)
if solvent == 'explicit':
thermodynamic_state = ThermodynamicState(system=systems[environment], temperature=temperature, pressure=pressure)
sampler_state = SamplerState(system=systems[environment], positions=positions[environment], box_vectors=systems[environment].getDefaultPeriodicBoxVectors())
else:
thermodynamic_state = ThermodynamicState(system=systems[environment], temperature=temperature)
sampler_state = SamplerState(system=systems[environment], positions=positions[environment])
mcmc_samplers[environment] = MCMCSampler(thermodynamic_state, sampler_state, storage=storage)
mcmc_samplers[environment].nsteps = 5 # reduce number of steps for testing
mcmc_samplers[environment].verbose = True
exen_samplers[environment] = ExpandedEnsembleSampler(mcmc_samplers[environment], topologies[environment], chemical_state_key, proposal_engines[environment], options={'nsteps':10}, geometry_engine=geometry_engine, storage=storage)
exen_samplers[environment].verbose = True
sams_samplers[environment] = SAMSSampler(exen_samplers[environment], storage=storage)
sams_samplers[environment].verbose = True
thermodynamic_states[environment] = thermodynamic_state
# Create test MultiTargetDesign sampler.
from perses.samplers.samplers import MultiTargetDesign
target_samplers = {sams_samplers['explicit-src-imatinib']: 1.0, sams_samplers['explicit-abl-imatinib']: -1.0}
designer = MultiTargetDesign(target_samplers, storage=self.storage)
designer.verbose = True
# Store things.
self.molecules = molecules
self.environments = environments
self.topologies = topologies
self.positions = positions
self.system_generators = system_generators
self.systems = systems
self.proposal_engines = proposal_engines
self.thermodynamic_states = thermodynamic_states
self.mcmc_samplers = mcmc_samplers
self.exen_samplers = exen_samplers
self.sams_samplers = sams_samplers
self.designer = designer
# This system must currently be minimized.
minimize(self)
class ProtonationStateSampler(object):
"""
Protonation state sampler with given fixed target probabilities for ligand in solvent.
Parameters
----------
samplers : list of SAMSSampler
The SAMS samplers whose relative partition functions go into the design objective computation.
sampler_exponents : dict of SAMSSampler : float
samplers.keys() are the samplers, and samplers[key]
log_target_probabilities : dict of hashable object : float
log_target_probabilities[key] is the computed log objective function (target probability) for chemical state `key`
verbose : bool
If True, verbose output is printed.
"""
def __init__(self, complex_sampler, solvent_sampler, log_state_penalties, storage=None, verbose=False):
"""
Initialize a protonation state sampler with fixed target probabilities for ligand in solvent.
Parameters
----------
complex_sampler : ExpandedEnsembleSampler
Ligand in complex sampler
solvent_sampler : SAMSSampler
Ligand in solution sampler
log_state_penalties : dict
log_state_penalties[smiles] is the log state free energy (in kT) for ligand state 'smiles'
storage : NetCDFStorage, optional, default=None
If specified, will use the storage layer to write trajectory data.
verbose : bool, optional, default=False
If true, will print verbose output
"""
# Store target samplers.
self.log_state_penalties = log_state_penalties
self.samplers = [complex_sampler, solvent_sampler]
self.complex_sampler = complex_sampler
self.solvent_sampler = solvent_sampler
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize storage for target probabilities.
self.log_target_probabilities = { key : - log_state_penalties[key] for key in log_state_penalties }
self.verbose = verbose
self.iteration = 0
@property
def state_keys(self):
return self.log_target_probabilities.keys()
def update_samplers(self):
"""
Update all samplers.
"""
for sampler in self.samplers:
sampler.update()
def update_target_probabilities(self):
"""
Update all target probabilities.
"""
# Update the complex sampler log weights using the solvent sampler log weights
for key in self.solvent_sampler.state_keys:
self.complex_sampler.log_weights[key] = self.solvent_sampler.sampler.log_weights[key]
if self.verbose:
print("log_weights = %s" % str(self.solvent_sampler.sampler.log_weights))
def update(self):
"""
Run one iteration of the sampler.
"""
if self.verbose:
print("*" * 80)
print("ProtonationStateSampler iteration %8d" % self.iteration)
self.update_samplers()
self.update_target_probabilities()
if self.storage: self.storage.sync()
self.iteration += 1
if self.verbose:
print("*" * 80)
def run(self, niterations=1):
"""
Run the protonation state sampler for the specified number of iterations.
Parameters
----------
niterations : int
The number of iterations to run the sampler for.
"""
# Update all samplers.
for iteration in range(niterations):
self.update()
def __init__(self, sampler, topology, state_key, proposal_engine, geometry_engine, log_weights=None, options=None, platform=None, envname=None, storage=None, ncmc_write_interval=1):
"""
Create an expanded ensemble sampler.
p(x,k) \propto \exp[-u_k(x) + g_k]
where g_k is the log weight.
Parameters
----------
sampler : MCMCSampler
MCMCSampler initialized with current SamplerState
topology : simtk.openmm.app.Topology
Current topology
state : hashable object
Current chemical state
proposal_engine : ProposalEngine
ProposalEngine to use for proposing new chemical states
geometry_engine : GeometryEngine
GeometryEngine to use for dimension matching
log_weights : dict of object : float
Log weights to use for expanded ensemble biases.
options : dict, optional, default=dict()
Options for initializing switching scheme, such as 'timestep', 'nsteps', 'functions' for NCMC
platform : simtk.openmm.Platform, optional, default=None
Platform to use for NCMC switching. If `None`, default (fastest) platform is used.
storage : NetCDFStorageView, optional, default=None
If specified, use this storage layer.
ncmc_write_interval : int, default 1
How frequently to write out NCMC protocol steps.
"""
# Keep copies of initializing arguments.
# TODO: Make deep copies?
self.sampler = sampler
self._pressure = sampler.thermodynamic_state.pressure
self._temperature = sampler.thermodynamic_state.temperature
self._omm_topology = topology
self.topology = md.Topology.from_openmm(topology)
self.state_key = state_key
self.proposal_engine = proposal_engine
self.log_weights = log_weights
if self.log_weights is None: self.log_weights = dict()
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize
self.iteration = 0
option_names = ['timestep', 'nsteps', 'functions', 'nsteps_mcmc', 'splitting']
if options is None:
options = dict()
for option_name in option_names:
if option_name not in options:
options[option_name] = None
if options['splitting']:
self._ncmc_splitting = options['splitting']
else:
self._ncmc_splitting = "V R O H R V"
if options['nsteps']:
self._switching_nsteps = options['nsteps']
self.ncmc_engine = NCMCEngine(temperature=self.sampler.thermodynamic_state.temperature,
timestep=options['timestep'], nsteps=options['nsteps'],
functions=options['functions'], integrator_splitting=self._ncmc_splitting,
platform=platform, storage=self.storage,
write_ncmc_interval=ncmc_write_interval)
else:
self._switching_nsteps = 0
if options['nsteps_mcmc']:
self._n_iterations_per_update = options['nsteps_mcmc']
else:
self._n_iterations_per_update = 100
self.geometry_engine = geometry_engine
self.naccepted = 0
self.nrejected = 0
self.number_of_state_visits = dict()
self.verbose = False
self.pdbfile = None # if not None, write PDB file
self.geometry_pdbfile = None # if not None, write PDB file of geometry proposals
self.accept_everything = False # if True, will accept anything that doesn't lead to NaNs
self.logPs = list()
self.sampler.minimize(max_iterations=40)
modeller.addSolvent(system_generators[solvent].getForceField(), model='tip3p', padding=9.0*unit.angstrom)
topologies[environment] = modeller.getTopology()
positions[environment] = modeller.getPositions()
else:
environment = solvent + '-' + component
topologies[environment] = topologies[component]
positions[environment] = positions[component]
# Set up the proposal engines.
from perses.rjmc.topology_proposal import SmallMoleculeSetProposalEngine
proposal_metadata = { }
proposal_engines = dict()
for environment in environments:
storage = None
if self.storage:
storage = NetCDFStorageView(self.storage, envname=environment)
proposal_engines[environment] = SmallMoleculeSetProposalEngine(molecules, system_generators[environment], residue_name='MOL', storage=storage)
# Generate systems
systems = dict()
for environment in environments:
systems[environment] = system_generators[environment].build_system(topologies[environment])
# Define thermodynamic state of interest.
from perses.samplers.thermodynamics import ThermodynamicState
thermodynamic_states = dict()
temperature = 300*unit.kelvin
pressure = 1.0*unit.atmospheres
for component in components:
for solvent in solvents:
environment = solvent + '-' + component
def __init__(self, molecules: List[str], output_filename: str, ncmc_switching_times: Dict[str, int], equilibrium_steps: Dict[str, int], timestep: unit.Quantity, initial_molecule: str=None, geometry_options: Dict=None):
self._molecules = [SmallMoleculeSetProposalEngine.canonicalize_smiles(molecule) for molecule in molecules]
environments = ['explicit', 'vacuum']
temperature = 298.15 * unit.kelvin
pressure = 1.0 * unit.atmospheres
constraints = app.HBonds
self._storage = NetCDFStorage(output_filename)
self._ncmc_switching_times = ncmc_switching_times
self._n_equilibrium_steps = equilibrium_steps
self._geometry_options = geometry_options
# Create a system generator for our desired forcefields.
from perses.rjmc.topology_proposal import SystemGenerator
system_generators = dict()
from pkg_resources import resource_filename
gaff_xml_filename = resource_filename('perses', 'data/gaff.xml')
barostat = openmm.MonteCarloBarostat(pressure, temperature)
system_generators['explicit'] = SystemGenerator([gaff_xml_filename, 'tip3p.xml'],
forcefield_kwargs={'nonbondedCutoff': 9.0 * unit.angstrom,
'implicitSolvent': None,
'constraints': constraints,
'ewaldErrorTolerance': 1e-5,
'hydrogenMass': 3.0*unit.amu}, periodic_forcefield_kwargs = {'nonbondedMethod': app.PME}
barostat=barostat)
system_generators['vacuum'] = SystemGenerator([gaff_xml_filename],
forcefield_kwargs={'implicitSolvent': None,
'constraints': constraints,
'hydrogenMass': 3.0*unit.amu}, nonperiodic_forcefield_kwargs = {'nonbondedMethod': app.NoCutoff})
#
# Create topologies and positions
#
topologies = dict()
positions = dict()
from openmoltools import forcefield_generators
forcefield = app.ForceField(gaff_xml_filename, 'tip3p.xml')
forcefield.registerTemplateGenerator(forcefield_generators.gaffTemplateGenerator)
# Create molecule in vacuum.
from perses.utils.openeye import extractPositionsFromOEMol
from openmoltools.openeye import smiles_to_oemol, generate_conformers
if initial_molecule:
smiles = initial_molecule
else:
smiles = np.random.choice(molecules)
molecule = smiles_to_oemol(smiles)
molecule = generate_conformers(molecule, max_confs=1)
topologies['vacuum'] = forcefield_generators.generateTopologyFromOEMol(molecule)
positions['vacuum'] = extractPositionsFromOEMol(molecule)
# Create molecule in solvent.
modeller = app.Modeller(topologies['vacuum'], positions['vacuum'])
modeller.addSolvent(forcefield, model='tip3p', padding=9.0 * unit.angstrom)
topologies['explicit'] = modeller.getTopology()
positions['explicit'] = modeller.getPositions()
# Set up the proposal engines.
proposal_metadata = {}
proposal_engines = dict()
for environment in environments:
proposal_engines[environment] = SmallMoleculeSetProposalEngine(self._molecules,
system_generators[environment])
# Generate systems
systems = dict()
for environment in environments:
systems[environment] = system_generators[environment].build_system(topologies[environment])
# Define thermodynamic state of interest.
thermodynamic_states = dict()
thermodynamic_states['explicit'] = states.ThermodynamicState(system=systems['explicit'],
temperature=temperature, pressure=pressure)
thermodynamic_states['vacuum'] = states.ThermodynamicState(system=systems['vacuum'], temperature=temperature)
# Create SAMS samplers
from perses.samplers.samplers import ExpandedEnsembleSampler, SAMSSampler
mcmc_samplers = dict()
exen_samplers = dict()
sams_samplers = dict()
for environment in environments:
storage = NetCDFStorageView(self._storage, envname=environment)
if self._geometry_options:
n_torsion_divisions = self._geometry_options['n_torsion_divsions'][environment]
use_sterics = self._geometry_options['use_sterics'][environment]
else:
n_torsion_divisions = 180
use_sterics = False
geometry_engine = geometry.FFAllAngleGeometryEngine(storage=storage, n_torsion_divisions=n_torsion_divisions, use_sterics=use_sterics)
move = mcmc.LangevinSplittingDynamicsMove(timestep=timestep, splitting="V R O R V",
n_restart_attempts=10)
chemical_state_key = proposal_engines[environment].compute_state_key(topologies[environment])
if environment == 'explicit':
sampler_state = states.SamplerState(positions=positions[environment],
box_vectors=systems[environment].getDefaultPeriodicBoxVectors())
else:
sampler_state = states.SamplerState(positions=positions[environment])
mcmc_samplers[environment] = mcmc.MCMCSampler(thermodynamic_states[environment], sampler_state, move)
exen_samplers[environment] = ExpandedEnsembleSampler(mcmc_samplers[environment], topologies[environment],
chemical_state_key, proposal_engines[environment],
geometry_engine,
options={'nsteps': self._ncmc_switching_times[environment]}, storage=storage, ncmc_write_interval=self._ncmc_switching_times[environment])
exen_samplers[environment].verbose = True
sams_samplers[environment] = SAMSSampler(exen_samplers[environment], storage=storage)
sams_samplers[environment].verbose = True
# Create test MultiTargetDesign sampler.
from perses.samplers.samplers import MultiTargetDesign
target_samplers = {sams_samplers['explicit']: 1.0, sams_samplers['vacuum']: -1.0}
designer = MultiTargetDesign(target_samplers, storage=self._storage)
# Store things.
self.molecules = molecules
self.environments = environments
self.topologies = topologies
self.positions = positions
self.system_generators = system_generators
self.proposal_engines = proposal_engines
self.thermodynamic_states = thermodynamic_states
self.mcmc_samplers = mcmc_samplers
self.exen_samplers = exen_samplers
self.sams_samplers = sams_samplers
self.designer = designer
class NCMCEngine(object):
"""
NCMC switching engine
Examples
--------
Create a transformation for an alanine dipeptide test system where the N-methyl group is eliminated.
>>> from openmmtools import testsystems
>>> testsystem = testsystems.AlanineDipeptideVacuum()
>>> from perses.rjmc.topology_proposal import TopologyProposal
>>> new_to_old_atom_map = { index : index for index in range(testsystem.system.getNumParticles()) if (index > 3) } # all atoms but N-methyl
>>> topology_proposal = TopologyProposal(old_system=testsystem.system, old_topology=testsystem.topology, old_chemical_state_key='AA', new_chemical_state_key='AA', new_system=testsystem.system, new_topology=testsystem.topology, logp_proposal=0.0, new_to_old_atom_map=new_to_old_atom_map, metadata=dict())
>>> ncmc_engine = NCMCEngine(temperature=300.0*unit.kelvin, functions=default_functions, nsteps=50, timestep=1.0*unit.femtoseconds)
>>> positions = testsystem.positions
>>> [positions, logP_delete, potential_delete] = ncmc_engine.integrate(topology_proposal, positions, direction='delete')
>>> [positions, logP_insert, potential_insert] = ncmc_engine.integrate(topology_proposal, positions, direction='insert')
"""
def __init__(self,
temperature=default_temperature,
functions=None,
nsteps=default_nsteps,
steps_per_propagation=default_steps_per_propagation,
timestep=default_timestep,
constraint_tolerance=None,
platform=None,
write_ncmc_interval=1,
measure_shadow_work=False,
integrator_splitting='V R O H R V',
storage=None,
verbose=False,
LRUCapacity=10,
pressure=None,
bond_softening_constant=1.0,
angle_softening_constant=1.0):
"""
This is the base class for NCMC switching between two different systems.
Parameters
----------
temperature : simtk.unit.Quantity with units compatible with kelvin
The temperature at which switching is to be run
functions : dict of str:str, optional, default=default_functions
functions[parameter] is the function (parameterized by 't' which switched from 0 to 1) that
controls how alchemical context parameter 'parameter' is switched
nsteps : int, optional, default=1
The number of steps to use for switching.
steps_per_propagation : int, optional, default=1
The number of intermediate propagation steps taken at each switching step
timestep : simtk.unit.Quantity with units compatible with femtoseconds, optional, default=1*femtosecond
The timestep to use for integration of switching velocity Verlet steps.
constraint_tolerance : float, optional, default=None
If not None, this relative constraint tolerance is used for position and velocity constraints.
platform : simtk.openmm.Platform, optional, default=None
If specified, the platform to use for OpenMM simulations.
write_ncmc_interval : int, optional, default=None
If a positive integer is specified, a snapshot frame will be written to storage with the specified interval on NCMC switching.
'storage' must also be specified.
measure_shadow_work : bool, optional, default False
Whether to measure shadow work
integrator_splitting : str, optional, default='V R O H R V'
NCMC internal integrator splitting based on OpenMMTools Langevin splittings
storage : NetCDFStorageView, optional, default=None
If specified, write data using this class.
verbose : bool, optional, default=False
If True, print debug information.
LRUCapacity : int, default 10
Capacity of LRU cache for hybrid systems
pressure : float, default None
The pressure to use for the simulation. If None, no barostat
"""
# Handle some defaults.
if functions == None:
functions = LambdaProtocol.default_functions
if nsteps == None:
nsteps = default_nsteps
if timestep == None:
timestep = default_timestep
if temperature == None:
temperature = default_temperature
self._temperature = temperature
self._functions = copy.deepcopy(functions)
self._nsteps = nsteps
self._timestep = timestep
self._constraint_tolerance = constraint_tolerance
self._platform = platform
self._integrator_splitting = integrator_splitting
self._steps_per_propagation = steps_per_propagation
self._verbose = verbose
self._pressure = pressure
self._bond_softening_constant = bond_softening_constant
self._angle_softening_constant = angle_softening_constant
self._disable_barostat = False
self._hybrid_cache = LRUCache(capacity=LRUCapacity)
self._measure_shadow_work = measure_shadow_work
self._nattempted = 0
self._storage = None
if storage is not None:
self._storage = NetCDFStorageView(storage,
modname=self.__class__.__name__)
self._save_configuration = True
else:
self._save_configuration = False
if write_ncmc_interval is not None:
self._write_ncmc_interval = write_ncmc_interval
else:
self._write_ncmc_interval = 1
self._work_save_interval = write_ncmc_interval
@property
def beta(self):
kT = kB * self._temperature
beta = 1.0 / kT
return beta
def _compute_energy_contribution(self, hybrid_thermodynamic_state,
initial_sampler_state,
final_sampler_state):
"""
Compute NCMC energy contribution to log probability.
See Eqs. 62 and 63 (two-stage) and Eq. 45 (hybrid) of reference document.
In both cases, the contribution is u(final_positions, final_lambda) - u(initial_positions, initial_lambda).
Parameters
----------
hybrid_thermodynamic_state : openmmtools.states.CompoundThermodynamicState
The thermodynamic state of the hybrid sampler.
initial_sampler_state : openmmtools.states.SamplerState
The sampler state of the nonalchemical system at the start of the NCMC protocol with box vectors
final_sampler_state : openmmtools.states.SamplerState
The sampler state of the nonalchemical system at the end of the NCMC protocol
Returns
-------
logP_energy : float
The NCMC energy contribution to log probability.
"""
hybrid_thermodynamic_state.set_alchemical_parameters(0.0)
initial_reduced_potential = compute_reduced_potential(
hybrid_thermodynamic_state, initial_sampler_state)
hybrid_thermodynamic_state.set_alchemical_parameters(1.0)
final_reduced_potential = compute_reduced_potential(
hybrid_thermodynamic_state, final_sampler_state)
return final_reduced_potential - initial_reduced_potential
def _topology_proposal_to_thermodynamic_states(self, topology_proposal):
"""
Convert a topology proposal to thermodynamic states for the end systems. This will be used to compute the
"logP_energy" quantity.
Parameters
----------
topology_proposal : perses.rjmc.TopologyProposal
topology proposal for whose endpoint systems we want ThermodynamicStates
Returns
-------
old_thermodynamic_state : openmmtools.states.ThermodynamicState
The old system (nonalchemical) thermodynamic state
new_thermodynamic_state : openmmtools.states.ThermodynamicState
The new system (nonalchemical) thermodynamic state
"""
systems = [topology_proposal.old_system, topology_proposal.new_system]
thermostates = []
for system in systems:
thermodynamic_state = ThermodynamicState(
system, temperature=self._temperature, pressure=self._pressure)
thermostates.append(thermodynamic_state)
return thermostates[0], thermostates[1]
def make_alchemical_system(self, topology_proposal, current_positions,
new_positions):
"""
Generate an alchemically-modified system at the correct atoms
based on the topology proposal. This method generates a hybrid system using the new
HybridTopologyFactory. It memoizes so that calling multiple times (within a recent time period)
will immediately return a cached object.
Parameters
----------
topology_proposal : perses.rjmc.TopologyProposal
Unmodified real system corresponding to appropriate leg of transformation.
current_positions : np.ndarray of float
Positions of "old" system
new_positions : np.ndarray of float
Positions of "new" system atoms
Returns
-------
hybrid_factory : perses.annihilation.relative.HybridTopologyFactory
a factory object containing the hybrid system
"""
try:
hybrid_factory = self._hybrid_cache[topology_proposal]
#If we've retrieved the factory from the cache, update it to include the relevant positions
hybrid_factory._old_positions = current_positions
hybrid_factory._new_positions = new_positions
hybrid_factory._compute_hybrid_positions()
except KeyError:
try:
hybrid_factory = HybridTopologyFactory(
topology_proposal,
current_positions,
new_positions,
bond_softening_constant=self._bond_softening_constant,
angle_softening_constant=self._angle_softening_constant)
self._hybrid_cache[topology_proposal] = hybrid_factory
except:
hybrid_factory = None
return hybrid_factory
def integrate(self,
topology_proposal,
initial_sampler_state,
proposed_sampler_state,
iteration=None):
"""
Performs NCMC switching to either delete or insert atoms according to the provided `topology_proposal`.
For `delete`, the system is first modified from fully interacting to alchemically modified, and then NCMC switching is used to eliminate atoms.
For `insert`, the system begins with eliminated atoms in an alchemically noninteracting form and NCMC switching is used to turn atoms on, followed by making system real.
Parameters
----------
topology_proposal : TopologyProposal
Contains old/new Topology and System objects and atom mappings.
initial_sampler_state : openmmtools.states.SamplerState representing the initial (old) system
Configurational properties of the atoms at the beginning of the NCMC switching.
proposed_sampler_state : openmmtools.states.SamplerState representing the proposed (post-geometry new) system
Configurational properties new system atoms at beginning of NCMC switching
iteration : int, optional, default=None
Iteration number, for storage purposes.
Returns
-------
final_old_sampler_state : openmmtools.State.SamplerState
The final configurational properties of the old system after hybrid alchemical switching
final_sampler_state : openmmtools.states.SamplerState
The final configurational properties after `nsteps` steps of alchemical switching, and reversion to the nonalchemical system
logP_work : float
The NCMC work contribution to the log acceptance probability (Eqs. 62 and 63)
logP_initial : float
The initial logP of the hybrid configuration
logP_final : float
The final logP of the hybrid configuration
"""
assert not initial_sampler_state.has_nan(
) and not proposed_sampler_state.has_nan()
#generate or retrieve the hybrid topology factory:
hybrid_factory = self.make_alchemical_system(
topology_proposal, initial_sampler_state.positions,
proposed_sampler_state.positions)
if hybrid_factory is None:
_logger.warning(
"Unable to construct hybrid system for {} -> {}".format(
topology_proposal.old_chemical_state_key,
topology_proposal.new_chemical_state_key))
return initial_sampler_state, proposed_sampler_state, -np.inf, 0.0, 0.0
topology = hybrid_factory.hybrid_topology
#generate the corresponding thermodynamic and sampler states so that we can use the NonequilibriumSwitchingMove:
#First generate the thermodynamic state:
hybrid_system = hybrid_factory.hybrid_system
hybrid_thermodynamic_state = ThermodynamicState(
hybrid_system,
temperature=self._temperature,
pressure=self._pressure)
#Now create an RelativeAlchemicalState from the hybrid system:
alchemical_state = RelativeAlchemicalState.from_system(hybrid_system)
alchemical_state.set_alchemical_parameters(0.0)
#Now create a compound thermodynamic state that combines the hybrid thermodynamic state with the alchemical state:
compound_thermodynamic_state = CompoundThermodynamicState(
hybrid_thermodynamic_state, composable_states=[alchemical_state])
#construct a sampler state from the hybrid positions and the box vectors of the initial sampler state:
initial_hybrid_positions = hybrid_factory.hybrid_positions
initial_hybrid_box_vectors = initial_sampler_state.box_vectors
initial_hybrid_sampler_state = SamplerState(
initial_hybrid_positions, box_vectors=initial_hybrid_box_vectors)
final_hybrid_sampler_state = copy.deepcopy(
initial_hybrid_sampler_state)
#create the nonequilibrium move:
#ne_move = NonequilibriumSwitchingMove(self._functions, self._integrator_splitting, self._temperature, self._nsteps, self._timestep,
# work_save_interval=self._write_ncmc_interval, top=topology,subset_atoms=None,
# save_configuration=self._save_configuration, measure_shadow_work=self._measure_shadow_work)
ne_move = ExternalNonequilibriumSwitchingMove(
self._functions,
nsteps_neq=self._nsteps,
timestep=self._timestep,
temperature=self._temperature,
work_configuration_save_interval=self._work_save_interval,
splitting="V R O R V")
#run the NCMC protocol
try:
ne_move.apply(compound_thermodynamic_state,
final_hybrid_sampler_state)
except Exception as e:
_logger.warn("NCMC failed because {}; rejecting.".format(str(e)))
logP_work = -np.inf
return [
initial_sampler_state, proposed_sampler_state, -np.inf, 0.0,
0.0
]
#get the total work:
logP_work = -ne_move.cumulative_work[-1]
# Compute contribution of transforming to and from the hybrid system:
context, integrator = global_context_cache.get_context(
hybrid_thermodynamic_state)
#set all alchemical parameters to zero:
for parameter in self._functions.keys():
context.setParameter(parameter, 0.0)
initial_hybrid_sampler_state.apply_to_context(context,
ignore_velocities=True)
initial_reduced_potential = hybrid_thermodynamic_state.reduced_potential(
context)
#set all alchemical parameters to one:
for parameter in self._functions.keys():
context.setParameter(parameter, 1.0)
final_hybrid_sampler_state.apply_to_context(context,
ignore_velocities=True)
final_reduced_potential = hybrid_thermodynamic_state.reduced_potential(
context)
#reset the parameters back to zero just in case
for parameter in self._functions.keys():
context.setParameter(parameter, 0.0)
#compute the output SamplerState, which has the atoms only for the new system post-NCMC:
new_positions = hybrid_factory.new_positions(
final_hybrid_sampler_state.positions)
new_box_vectors = final_hybrid_sampler_state.box_vectors
final_sampler_state = SamplerState(new_positions,
box_vectors=new_box_vectors)
#compute the output SamplerState for the atoms only in the old system (required for geometry_logP_reverse)
old_positions = hybrid_factory.old_positions(
final_hybrid_sampler_state.positions)
old_box_vectors = copy.deepcopy(
new_box_vectors) #these are the same as the new system
final_old_sampler_state = SamplerState(old_positions,
box_vectors=old_box_vectors)
#extract the trajectory and box vectors from the move:
trajectory = ne_move.trajectory[::-self.
_write_ncmc_interval, :, :][::-1]
topology = hybrid_factory.hybrid_topology
position_varname = "ncmcpositions"
nframes = np.shape(trajectory)[0]
#extract box vectors:
box_vec_varname = "ncmcboxvectors"
box_lengths = ne_move.box_lengths[::-self.
_write_ncmc_interval, :][::-1]
box_angles = ne_move.box_angles[::-self._write_ncmc_interval, :][::-1]
box_lengths_and_angles = np.stack([box_lengths, box_angles])
#write out the positions of the topology
if self._storage:
for frame in range(nframes):
self._storage.write_configuration(position_varname,
trajectory[frame, :, :],
topology,
iteration=iteration,
frame=frame,
nframes=nframes)
#write out the periodict box vectors:
if self._storage:
self._storage.write_array(box_vec_varname,
box_lengths_and_angles,
iteration=iteration)
#retrieve the protocol work and write that out too:
protocol_work = ne_move.cumulative_work
if self._storage:
self._storage.write_array("protocolwork",
protocol_work,
iteration=iteration)
# Return
return [
final_old_sampler_state, final_sampler_state, logP_work,
-initial_reduced_potential, -final_reduced_potential
]
class SAMSSampler(object):
"""
Self-adjusted mixture sampling engine.
Properties
----------
state_keys : set of objects
The names of states sampled by the sampler.
logZ : dict() of keys : float
logZ[key] is the log partition function (up to an additive constant) estimate for chemical state `key`
update_method : str
Update method. One of ['default']
iteration : int
Iterations completed.
verbose : bool
If True, verbose debug output is printed.
References
----------
[1] Tan, Z. (2015) Optimally adjusted mixture sampling and locally weighted histogram analysis, Journal of Computational and Graphical Statistics, to appear. (Supplement)
http://www.stat.rutgers.edu/home/ztan/Publication/SAMS_redo4.pdf
Examples
--------
>>> # Create a test system
>>> test = testsystems.AlanineDipeptideVacuum()
>>> # Create a SystemGenerator and rebuild the System.
>>> from perses.rjmc.topology_proposal import SystemGenerator
>>> system_generator = SystemGenerator(['amber99sbildn.xml'], forcefield_kwargs={ 'nonbondedMethod' : app.NoCutoff, 'implicitSolvent' : None, 'constraints' : None })
>>> test.system = system_generator.build_system(test.topology)
>>> # Create a sampler state.
>>> sampler_state = SamplerState(system=test.system, positions=test.positions)
>>> # Create a thermodynamic state.
>>> thermodynamic_state = ThermodynamicState(system=test.system, temperature=298.0*unit.kelvin)
>>> # Create an MCMC sampler
>>> mcmc_sampler = MCMCSampler(thermodynamic_state, sampler_state)
>>> # Turn off verbosity
>>> mcmc_sampler.verbose = False
>>> from perses.rjmc.geometry import FFAllAngleGeometryEngine
>>> geometry_engine = FFAllAngleGeometryEngine(metadata={})
>>> # Create an Expanded Ensemble sampler
>>> from perses.rjmc.topology_proposal import PointMutationEngine
>>> allowed_mutations = [[('2','ALA')],[('2','VAL'),('2','LEU')]]
>>> proposal_engine = PointMutationEngine(test.topology, system_generator, max_point_mutants=1, chain_id='1', proposal_metadata=None, allowed_mutations=allowed_mutations)
>>> exen_sampler = ExpandedEnsembleSampler(mcmc_sampler, test.topology, 'ACE-ALA-NME', proposal_engine, geometry_engine)
>>> # Create a SAMS sampler
>>> sams_sampler = SAMSSampler(exen_sampler)
>>> # Run the sampler
>>> sams_sampler.run() # doctest: +ELLIPSIS
...
"""
def __init__(self, sampler, logZ=None, log_target_probabilities=None, update_method='two-stage', storage=None, second_stage_start=1000):
"""
Create a SAMS Sampler.
Parameters
----------
sampler : ExpandedEnsembleSampler
The expanded ensemble sampler used to sample both configurations and discrete thermodynamic states.
logZ : dict of key : float, optional, default=None
If specified, the log partition functions for each state will be initialized to the specified dictionary.
log_target_probabilities : dict of key : float, optional, default=None
If specified, unnormalized target probabilities; default is all 0.
update_method : str, optional, default='default'
SAMS update algorithm
storage : NetCDFStorageView, optional, default=None
second_state_start : int, optional, default None
At what iteration number to switch to the optimal gain decay
"""
from scipy.misc import logsumexp
from perses.tests.utils import createOEMolFromSMILES
# Keep copies of initializing arguments.
# TODO: Make deep copies?
self.sampler = sampler
self.chemical_states = None
self._reference_state = None
try:
self.chemical_states = self.sampler.proposal_engine.chemical_state_list
except NotImplementedError:
logger.warn("The proposal engine has not properly implemented the chemical state property; SAMS will add states on the fly.")
if self.chemical_states:
# Select a reference state that will always be subtracted (ensure that dict ordering does not change)
self._reference_state = self.chemical_states[0]
# Initialize the logZ dictionary with scores based on the number of atoms
# This is not the negative because the weights are set to the negative of the initial weights
self.logZ = {chemical_state: self._num_dof_compensation(chemical_state) for chemical_state in self.chemical_states}
#Initialize log target probabilities with log(1/n_states)
self.log_target_probabilities = {chemical_state : np.log(len(self.chemical_states)) for chemical_state in self.chemical_states}
#If initial weights are specified, override any weight with what is provided
#However, if the chemical state is not in the reachable chemical state list,throw an exception
if logZ is not None:
for (chemical_state, logZ_value) in logZ:
if chemical_state not in self.chemical_states:
raise ValueError("Provided a logZ initial value for an un-proposable chemical state")
self.logZ[chemical_state] = logZ_value
if log_target_probabilities is not None:
for (chemical_state, log_target_probability) in log_target_probabilities:
if chemical_state not in self.chemical_states:
raise ValueError("Provided a log target probability for an un-proposable chemical state.")
self.log_target_probabilities[chemical_state] = log_target_probability
#normalize target probabilities
#this is likely not necessary, but it is copying the algorithm in Ref 1
log_sum_target_probabilities = logsumexp((list(self.log_target_probabilities.values())))
self.log_target_probabilities = {chemical_state : log_target_probability - log_sum_target_probabilities for chemical_state, log_target_probability in self.log_target_probabilities}
else:
self.logZ = dict()
self.log_target_probabilities = dict()
self.update_method = update_method
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize.
self.iteration = 0
self.verbose = False
self.sampler.log_weights = {state_key: - self.logZ[state_key] for state_key in self.logZ.keys()}
self.second_stage_start = 0
if second_stage_start is not None:
self.second_stage_start = second_stage_start
@property
def state_keys(self):
return self.logZ.keys()
def _num_dof_compensation(self, smiles):
"""
Compute an approximate compensating factor for a chemical state based on the number of degrees of freedom that it has.
The formula is:
(num_heavy*heavy_factor) + (num_hydrogen*h_factor) where
heavy_factor = 4.5 and
light_factor = 3.8
Parameters
----------
smiles : str
The SMILES string of the molecule
Returns
-------
correction_factor : float
"""
mol = createOEMolFromSMILES(smiles)
num_heavy = 0
num_light = 0
heavy_factor = 4.5
light_factor = 3.8
for atom in mol.GetAtoms():
if atom.GetAtomicNum() == 1:
num_light += 1
else:
num_heavy += 1
correction_factor = num_heavy*heavy_factor + num_light*light_factor
return correction_factor
def update_sampler(self):
"""
Update the underlying expanded ensembles sampler.
"""
self.sampler.update()
def update_logZ_estimates(self):
"""
Update the logZ estimates according to self.update_method.
"""
state_key = self.sampler.state_key
# Add state key to dictionaries if we haven't visited this state before.
if state_key not in self.logZ:
logger.warn("A new state key is being added to the logZ; note that this makes the resultant algorithm different from SAMS")
self.logZ[state_key] = 0.0
if state_key not in self.log_target_probabilities:
logger.warn("A new state key is being added to the target probabilities; note that this makes the resultant algorithm different from SAMS")
self.log_target_probabilities[state_key] = 0.0
# Update estimates of logZ.
if self.update_method == 'one-stage':
# Based on Eq. 9 of Ref. [1]
gamma = 1.0 / float(self.iteration+1)
elif self.update_method == 'two-stage':
# Keep gamma large until second stage is activated.
if self.iteration < self.second_stage_start:
# First stage.
gamma = 1.0
# TODO: Determine when to switch to second stage
else:
# Second stage.
gamma = 1.0 / float(self.iteration - self.second_stage_start + 1)
else:
raise Exception("SAMS update method '%s' unknown." % self.update_method)
#get the (t-1/2) update from equation 9 in ref 1
self.logZ[state_key] += gamma / np.exp(self.log_target_probabilities[state_key])
if self._reference_state:
#the second step of the (t-1/2 update), subtracting the reference state from everything else.
#we can only do this for cases where all states have been enumerated
self.logZ = {state_key : logZ_estimate - self.logZ[self._reference_state] for state_key, logZ_estimate in self.logZ.items()}
# Update log weights for sampler.
self.sampler.log_weights = { state_key : - self.logZ[state_key] for state_key in self.logZ.keys()}
if self.storage:
self.storage.write_object('logZ', self.logZ, iteration=self.iteration)
self.storage.write_object('log_weights', self.sampler.log_weights, iteration=self.iteration)
def update(self):
"""
Update the sampler with one step of sampling.
"""
if self.verbose:
print("=" * 80)
print("SAMS sampler iteration %5d" % self.iteration)
self.update_sampler()
self.update_logZ_estimates()
if self.storage: self.storage.sync()
self.iteration += 1
if self.verbose:
print("=" * 80)
def run(self, niterations=1):
"""
Run the sampler for the specified number of iterations
Parameters
----------
niterations : int, optional, default=1
Number of iterations to run the sampler for.
"""
for iteration in range(niterations):
self.update()
def __init__(self, sampler, logZ=None, log_target_probabilities=None, update_method='two-stage', storage=None, second_stage_start=1000):
"""
Create a SAMS Sampler.
Parameters
----------
sampler : ExpandedEnsembleSampler
The expanded ensemble sampler used to sample both configurations and discrete thermodynamic states.
logZ : dict of key : float, optional, default=None
If specified, the log partition functions for each state will be initialized to the specified dictionary.
log_target_probabilities : dict of key : float, optional, default=None
If specified, unnormalized target probabilities; default is all 0.
update_method : str, optional, default='default'
SAMS update algorithm
storage : NetCDFStorageView, optional, default=None
second_state_start : int, optional, default None
At what iteration number to switch to the optimal gain decay
"""
from scipy.special import logsumexp
from perses.utils.openeye import smiles_to_oemol
# Keep copies of initializing arguments.
# TODO: Make deep copies?
self.sampler = sampler
self.chemical_states = None
self._reference_state = None
try:
self.chemical_states = self.sampler.proposal_engine.chemical_state_list
except NotImplementedError:
_logger.warning("The proposal engine has not properly implemented the chemical state property; SAMS will add states on the fly.")
if self.chemical_states:
# Select a reference state that will always be subtracted (ensure that dict ordering does not change)
self._reference_state = self.chemical_states[0]
# Initialize the logZ dictionary with scores based on the number of atoms
# This is not the negative because the weights are set to the negative of the initial weights
self.logZ = {chemical_state: self._num_dof_compensation(chemical_state) for chemical_state in self.chemical_states}
#Initialize log target probabilities with log(1/n_states)
self.log_target_probabilities = {chemical_state : np.log(len(self.chemical_states)) for chemical_state in self.chemical_states}
#If initial weights are specified, override any weight with what is provided
#However, if the chemical state is not in the reachable chemical state list,throw an exception
if logZ is not None:
for (chemical_state, logZ_value) in logZ:
if chemical_state not in self.chemical_states:
raise ValueError("Provided a logZ initial value for an un-proposable chemical state")
self.logZ[chemical_state] = logZ_value
if log_target_probabilities is not None:
for (chemical_state, log_target_probability) in log_target_probabilities:
if chemical_state not in self.chemical_states:
raise ValueError("Provided a log target probability for an un-proposable chemical state.")
self.log_target_probabilities[chemical_state] = log_target_probability
#normalize target probabilities
#this is likely not necessary, but it is copying the algorithm in Ref 1
log_sum_target_probabilities = logsumexp((list(self.log_target_probabilities.values())))
self.log_target_probabilities = {chemical_state : log_target_probability - log_sum_target_probabilities for chemical_state, log_target_probability in self.log_target_probabilities}
else:
self.logZ = dict()
self.log_target_probabilities = dict()
self.update_method = update_method
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize.
self.iteration = 0
self.verbose = False
self.sampler.log_weights = {state_key: - self.logZ[state_key] for state_key in self.logZ.keys()}
self.second_stage_start = 0
if second_stage_start is not None:
self.second_stage_start = second_stage_start
def __init__(self, sampler, topology, state_key, proposal_engine, geometry_engine, log_weights=None, options=None, platform=None, envname=None, storage=None, ncmc_write_interval=1):
"""
Create an expanded ensemble sampler.
p(x,k) \propto \exp[-u_k(x) + g_k]
where g_k is the log weight.
Parameters
----------
sampler : MCMCSampler
MCMCSampler initialized with current SamplerState
topology : simtk.openmm.app.Topology
Current topology
state : hashable object
Current chemical state
proposal_engine : ProposalEngine
ProposalEngine to use for proposing new chemical states
geometry_engine : GeometryEngine
GeometryEngine to use for dimension matching
log_weights : dict of object : float
Log weights to use for expanded ensemble biases.
options : dict, optional, default=dict()
Options for initializing switching scheme, such as 'timestep', 'nsteps', 'functions' for NCMC
platform : simtk.openmm.Platform, optional, default=None
Platform to use for NCMC switching. If `None`, default (fastest) platform is used.
storage : NetCDFStorageView, optional, default=None
If specified, use this storage layer.
ncmc_write_interval : int, default 1
How frequently to write out NCMC protocol steps.
"""
# Keep copies of initializing arguments.
# TODO: Make deep copies?
self.sampler = sampler
self._pressure = sampler.thermodynamic_state.pressure
self._temperature = sampler.thermodynamic_state.temperature
self.topology = md.Topology.from_openmm(topology)
self.state_key = state_key
self.proposal_engine = proposal_engine
self.log_weights = log_weights
if self.log_weights is None: self.log_weights = dict()
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize
self.iteration = 0
option_names = ['timestep', 'nsteps', 'functions', 'nsteps_mcmc', 'splitting']
if options is None:
options = dict()
for option_name in option_names:
if option_name not in options:
options[option_name] = None
if options['splitting']:
self._ncmc_splitting = options['splitting']
else:
self._ncmc_splitting = "V R O H R V"
if options['nsteps']:
self._switching_nsteps = options['nsteps']
self.ncmc_engine = NCMCEngine(temperature=self.sampler.thermodynamic_state.temperature,
timestep=options['timestep'], nsteps=options['nsteps'],
functions=options['functions'], integrator_splitting=self._ncmc_splitting,
platform=platform, storage=self.storage,
write_ncmc_interval=ncmc_write_interval)
else:
self._switching_nsteps = 0
if options['nsteps_mcmc']:
self._n_iterations_per_update = options['nsteps_mcmc']
else:
self._n_iterations_per_update = 100
self.geometry_engine = geometry_engine
self.naccepted = 0
self.nrejected = 0
self.number_of_state_visits = dict()
self.verbose = False
self.pdbfile = None # if not None, write PDB file
self.geometry_pdbfile = None # if not None, write PDB file of geometry proposals
self.accept_everything = False # if True, will accept anything that doesn't lead to NaNs
self.logPs = list()
self.sampler.minimize(max_iterations=40)
class ExpandedEnsembleSampler(object):
"""
Method of expanded ensembles sampling engine.
The acceptance criteria is given in the reference document. Roughly, the proposal scheme is:
* Draw a proposed chemical state k', and calculate reverse proposal probability
* Conditioned on k' and the current positions x, generate new positions with the GeometryEngine
* With new positions, jump to a hybrid system at lambda=0
* Anneal from lambda=0 to lambda=1, accumulating work
* Jump from the hybrid system at lambda=1 to the k' system, and compute reverse GeometryEngine proposal
* Add weight of chemical states k and k' to acceptance probabilities
Properties
----------
sampler : MCMCSampler
The MCMC sampler used for updating positions.
proposal_engine : ProposalEngine
The ProposalEngine to use for proposing new sampler states and topologies.
system_generator : SystemGenerator
The SystemGenerator to use for creating System objects following proposals.
state : hashable object
The current sampler state. Can be any hashable object.
states : set of hashable object
All known states.
iteration : int
Iterations completed.
naccepted : int
Number of accepted thermodynamic/chemical state changes.
nrejected : int
Number of rejected thermodynamic/chemical state changes.
number_of_state_visits : dict of state_key
Cumulative counts of visited states.
verbose : bool
If True, verbose output is printed.
References
----------
[1] Lyubartsev AP, Martsinovski AA, Shevkunov SV, and Vorontsov-Velyaminov PN. New approach to Monte Carlo calculation of the free energy: Method of expanded ensembles. JCP 96:1776, 1992
http://dx.doi.org/10.1063/1.462133
Examples
--------
>>> # Create a test system
>>> test = testsystems.AlanineDipeptideVacuum()
>>> # Create a SystemGenerator and rebuild the System.
>>> from perses.rjmc.topology_proposal import SystemGenerator
>>> system_generator = SystemGenerator(['amber99sbildn.xml'], forcefield_kwargs={'implicitSolvent' : None, 'constraints' : None }, nonperiodic_forcefield_kwargs={'nonbondedMethod' : app.NoCutoff})
>>> test.system = system_generator.build_system(test.topology)
>>> # Create a sampler state.
>>> sampler_state = SamplerState(system=test.system, positions=test.positions)
>>> # Create a thermodynamic state.
>>> thermodynamic_state = ThermodynamicState(system=test.system, temperature=298.0*unit.kelvin)
>>> # Create an MCMC sampler
>>> mcmc_sampler = MCMCSampler(thermodynamic_state, sampler_state)
>>> # Turn off verbosity
>>> mcmc_sampler.verbose = False
>>> # Create an Expanded Ensemble sampler
>>> from perses.rjmc.topology_proposal import PointMutationEngine
>>> from perses.rjmc.geometry import FFAllAngleGeometryEngine
>>> geometry_engine = FFAllAngleGeometryEngine(metadata={})
>>> allowed_mutations = [[('2','ALA')],[('2','VAL'),('2','LEU')]]
>>> proposal_engine = PointMutationEngine(test.topology, system_generator, max_point_mutants=1, chain_id='1', proposal_metadata=None, allowed_mutations=allowed_mutations)
>>> exen_sampler = ExpandedEnsembleSampler(mcmc_sampler, test.topology, 'ACE-ALA-NME', proposal_engine, geometry_engine)
>>> # Run the sampler
>>> exen_sampler.run()
"""
def __init__(self, sampler, topology, state_key, proposal_engine, geometry_engine, log_weights=None, options=None, platform=None, envname=None, storage=None, ncmc_write_interval=1):
"""
Create an expanded ensemble sampler.
p(x,k) \propto \exp[-u_k(x) + g_k]
where g_k is the log weight.
Parameters
----------
sampler : MCMCSampler
MCMCSampler initialized with current SamplerState
topology : simtk.openmm.app.Topology
Current topology
state : hashable object
Current chemical state
proposal_engine : ProposalEngine
ProposalEngine to use for proposing new chemical states
geometry_engine : GeometryEngine
GeometryEngine to use for dimension matching
log_weights : dict of object : float
Log weights to use for expanded ensemble biases.
options : dict, optional, default=dict()
Options for initializing switching scheme, such as 'timestep', 'nsteps', 'functions' for NCMC
platform : simtk.openmm.Platform, optional, default=None
Platform to use for NCMC switching. If `None`, default (fastest) platform is used.
storage : NetCDFStorageView, optional, default=None
If specified, use this storage layer.
ncmc_write_interval : int, default 1
How frequently to write out NCMC protocol steps.
"""
# Keep copies of initializing arguments.
# TODO: Make deep copies?
self.sampler = sampler
self._pressure = sampler.thermodynamic_state.pressure
self._temperature = sampler.thermodynamic_state.temperature
self._omm_topology = topology
self.topology = md.Topology.from_openmm(topology)
self.state_key = state_key
self.proposal_engine = proposal_engine
self.log_weights = log_weights
if self.log_weights is None: self.log_weights = dict()
self.storage = None
if storage is not None:
self.storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
# Initialize
self.iteration = 0
option_names = ['timestep', 'nsteps', 'functions', 'nsteps_mcmc', 'splitting']
if options is None:
options = dict()
for option_name in option_names:
if option_name not in options:
options[option_name] = None
if options['splitting']:
self._ncmc_splitting = options['splitting']
else:
self._ncmc_splitting = "V R O H R V"
if options['nsteps']:
self._switching_nsteps = options['nsteps']
self.ncmc_engine = NCMCEngine(temperature=self.sampler.thermodynamic_state.temperature,
timestep=options['timestep'], nsteps=options['nsteps'],
functions=options['functions'], integrator_splitting=self._ncmc_splitting,
platform=platform, storage=self.storage,
write_ncmc_interval=ncmc_write_interval)
else:
self._switching_nsteps = 0
if options['nsteps_mcmc']:
self._n_iterations_per_update = options['nsteps_mcmc']
else:
self._n_iterations_per_update = 100
self.geometry_engine = geometry_engine
self.naccepted = 0
self.nrejected = 0
self.number_of_state_visits = dict()
self.verbose = False
self.pdbfile = None # if not None, write PDB file
self.geometry_pdbfile = None # if not None, write PDB file of geometry proposals
self.accept_everything = False # if True, will accept anything that doesn't lead to NaNs
self.logPs = list()
self.sampler.minimize(max_iterations=40)
@property
def state_keys(self):
return self.log_weights.keys()
def get_log_weight(self, state_key):
"""
Get the log weight of the specified state.
Parameters
----------
state_key : hashable object
The state key (e.g. chemical state key) to look up.
Returns
-------
log_weight : float
The log weight of the provided state key.
Notes
-----
This adds the key to the self.log_weights dict.
"""
if state_key not in self.log_weights:
self.log_weights[state_key] = 0.0
return self.log_weights[state_key]
def _system_to_thermodynamic_state(self, system):
"""
Given an OpenMM system object, create a corresponding ThermodynamicState that has the same
temperature and pressure as the current thermodynamic state.
Parameters
----------
system : openmm.System
The OpenMM system for which to create the thermodynamic state
Returns
-------
new_thermodynamic_state : openmmtools.states.ThermodynamicState
The thermodynamic state object representing the given system
"""
return ThermodynamicState(system, temperature=self._temperature, pressure=self._pressure)
def _geometry_forward(self, topology_proposal, old_sampler_state):
"""
Run geometry engine to propose new positions and compute logP
Parameters
----------
topology_proposal : TopologyProposal
Contains old/new Topology and System objects and atom mappings.
old_sampler_state : openmmtools.states.SamplerState
Configurational properties of the old system atoms.
Returns
-------
new_sampler_state : openmmtools.states.SamplerState
Configurational properties of new atoms proposed by geometry engine calculation.
geometry_logp_propose : float
The log probability of the forward-only proposal
"""
if self.verbose: print("Geometry engine proposal...")
# Generate coordinates for new atoms and compute probability ratio of old and new probabilities.
initial_time = time.time()
new_positions, geometry_logp_propose = self.geometry_engine.propose(topology_proposal, old_sampler_state.positions, self.sampler.thermodynamic_state.beta)
if self.verbose: print('proposal took %.3f s' % (time.time() - initial_time))
if self.geometry_pdbfile is not None:
print("Writing proposed geometry...")
from simtk.openmm.app import PDBFile
PDBFile.writeFile(topology_proposal.new_topology, new_positions, file=self.geometry_pdbfile)
self.geometry_pdbfile.flush()
new_sampler_state = SamplerState(new_positions, box_vectors=old_sampler_state.box_vectors)
return new_sampler_state, geometry_logp_propose
def _geometry_reverse(self, topology_proposal, new_sampler_state, old_sampler_state):
"""
Run geometry engine reverse calculation to determine logP
of proposing the old positions based on the new positions
Parameters
----------
topology_proposal : TopologyProposal
Contains old/new Topology and System objects and atom mappings.
new_sampler_state : openmmtools.states.SamplerState
Configurational properties of the new atoms.
old_sampler_state : openmmtools.states.SamplerState
Configurational properties of the old atoms.
Returns
-------
geometry_logp_reverse : float
The log probability of the proposal for the given transformation
"""
if self.verbose: print("Geometry engine logP_reverse calculation...")
initial_time = time.time()
geometry_logp_reverse = self.geometry_engine.logp_reverse(topology_proposal, new_sampler_state.positions, old_sampler_state.positions, self.sampler.thermodynamic_state.beta)
if self.verbose: print('calculation took %.3f s' % (time.time() - initial_time))
return geometry_logp_reverse
def _ncmc_hybrid(self, topology_proposal, old_sampler_state, new_sampler_state):
"""
Run a hybrid NCMC protocol from lambda = 0 to lambda = 1
Parameters
----------
topology_proposal : TopologyProposal
Contains old/new Topology and System objects and atom mappings.
old_sampler_State : openmmtools.states.SamplerState
SamplerState of old system at the beginning of NCMCSwitching
new_sampler_state : openmmtools.states.SamplerState
SamplerState of new system at the beginning of NCMCSwitching
Returns
-------
old_final_sampler_state : openmmtools.states.SamplerState
SamplerState of old system at the end of switching
new_final_sampler_state : openmmtools.states.SamplerState
SamplerState of new system at the end of switching
logP_work : float
The NCMC work contribution to the log acceptance probability (Eq. 44)
logP_energy : float
The contribution of switching to and from the hybrid system to the acceptance probability (Eq. 45)
"""
if self.verbose: print("Performing NCMC switching")
initial_time = time.time()
[ncmc_old_sampler_state, ncmc_new_sampler_state, logP_work, logP_initial_hybrid, logP_final_hybrid] = self.ncmc_engine.integrate(topology_proposal, old_sampler_state, new_sampler_state, iteration=self.iteration)
if self.verbose: print('NCMC took %.3f s' % (time.time() - initial_time))
# Check that positions are not NaN
if new_sampler_state.has_nan():
raise Exception("Positions are NaN after NCMC insert with %d steps" % self._switching_nsteps)
return ncmc_old_sampler_state, ncmc_new_sampler_state, logP_work, logP_initial_hybrid, logP_final_hybrid
def _geometry_ncmc_geometry(self, topology_proposal, sampler_state, old_log_weight, new_log_weight):
"""
Use a hybrid NCMC protocol to switch from the old system to new system
Will calculate new positions for the new system first, then give both
sets of positions to the hybrid NCMC integrator, and finally use the
final positions of the old and new systems to calculate the reverse
geometry probability
Parameters
----------
topology_proposal : TopologyProposal
Contains old/new Topology and System objects and atom mappings.
sampler_state : openmmtools.states.SamplerState
Configurational properties of old atoms at the beginning of the NCMC switching.
old_log_weight : float
Chemical state weight from SAMSSampler
new_log_weight : float
Chemical state weight from SAMSSampler
Returns
-------
logP_accept : float
Log of acceptance probability of entire Expanded Ensemble switch (Eq. 25 or 46)
ncmc_new_sampler_state : openmmtools.states.SamplerState
Configurational properties of new atoms at the end of the NCMC switching.
"""
if self.verbose: print("Updating chemical state with geometry-ncmc-geometry scheme...")
from perses.tests.utils import compute_potential
logP_chemical_proposal = topology_proposal.logp_proposal
old_thermodynamic_state = self.sampler.thermodynamic_state
new_thermodynamic_state = self._system_to_thermodynamic_state(topology_proposal.new_system)
initial_reduced_potential = feptasks.compute_reduced_potential(old_thermodynamic_state, sampler_state)
logP_initial_nonalchemical = - initial_reduced_potential
new_geometry_sampler_state, logP_geometry_forward = self._geometry_forward(topology_proposal, sampler_state)
#if we aren't doing any switching, then skip running the NCMC engine at all.
if self._switching_nsteps == 0:
ncmc_old_sampler_state = sampler_state
ncmc_new_sampler_state = new_geometry_sampler_state
logP_work = 0.0
logP_initial_hybrid = 0.0
logP_final_hybrid = 0.0
else:
ncmc_old_sampler_state, ncmc_new_sampler_state, logP_work, logP_initial_hybrid, logP_final_hybrid = self._ncmc_hybrid(topology_proposal, sampler_state, new_geometry_sampler_state)
if logP_work > -np.inf and logP_initial_hybrid > -np.inf and logP_final_hybrid > -np.inf:
logP_geometry_reverse = self._geometry_reverse(topology_proposal, ncmc_new_sampler_state, ncmc_old_sampler_state)
logP_to_hybrid = logP_initial_hybrid - logP_initial_nonalchemical
final_reduced_potential = feptasks.compute_reduced_potential(new_thermodynamic_state, ncmc_new_sampler_state)
logP_final_nonalchemical = -final_reduced_potential
logP_from_hybrid = logP_final_nonalchemical - logP_final_hybrid
logP_sams_weight = new_log_weight - old_log_weight
# Compute total log acceptance probability according to Eq. 46
logP_accept = logP_to_hybrid - logP_geometry_forward + logP_work + logP_from_hybrid + logP_geometry_reverse + logP_sams_weight
else:
logP_geometry_reverse = 0.0
logP_final = 0.0
logP_to_hybrid = 0.0
logP_from_hybrid = 0.0
logP_sams_weight = new_log_weight - old_log_weight
logP_accept = logP_to_hybrid - logP_geometry_forward + logP_work + logP_from_hybrid + logP_geometry_reverse + logP_sams_weight
#TODO: mark failed proposals as unproposable
if self.verbose:
print("logP_accept = %+10.4e [logP_to_hybrid = %+10.4e, logP_chemical_proposal = %10.4e, logP_reverse = %+10.4e, -logP_forward = %+10.4e, logP_work = %+10.4e, logP_from_hybrid = %+10.4e, logP_sams_weight = %+10.4e]"
% (logP_accept, logP_to_hybrid, logP_chemical_proposal, logP_geometry_reverse, -logP_geometry_forward, logP_work, logP_from_hybrid, logP_sams_weight))
# Write to storage.
if self.storage:
self.storage.write_quantity('logP_accept', logP_accept, iteration=self.iteration)
# Write components to storage
self.storage.write_quantity('logP_ncmc_work', logP_work, iteration=self.iteration)
self.storage.write_quantity('logP_from_hybrid', logP_from_hybrid, iteration=self.iteration)
self.storage.write_quantity('logP_to_hybrid', logP_to_hybrid, iteration=self.iteration)
self.storage.write_quantity('logP_chemical_proposal', logP_chemical_proposal, iteration=self.iteration)
self.storage.write_quantity('logP_reverse', logP_geometry_reverse, iteration=self.iteration)
self.storage.write_quantity('logP_forward', logP_geometry_forward, iteration=self.iteration)
self.storage.write_quantity('logP_sams_weight', logP_sams_weight, iteration=self.iteration)
# Write some aggregate statistics to storage to make contributions to acceptance probability easier to analyze
self.storage.write_quantity('logP_groups_chemical', logP_chemical_proposal, iteration=self.iteration)
self.storage.write_quantity('logP_groups_geometry', logP_geometry_reverse - logP_geometry_forward, iteration=self.iteration)
return logP_accept, ncmc_new_sampler_state
def update_positions(self, n_iterations=1):
"""
Sample new positions.
"""
self.sampler.run(n_iterations=n_iterations)
def update_state(self):
"""
Sample the thermodynamic state.
"""
initial_time = time.time()
# Propose new chemical state.
if self.verbose: print("Proposing new topology...")
[system, positions] = [self.sampler.thermodynamic_state.get_system(remove_thermostat=True), self.sampler.sampler_state.positions]
#omm_topology = topology.to_openmm() #convert to OpenMM topology for proposal engine
self._omm_topology.setPeriodicBoxVectors(self.sampler.sampler_state.box_vectors) #set the box vectors because in OpenMM topology has these...
topology_proposal = self.proposal_engine.propose(system, self._omm_topology)
if self.verbose: print("Proposed transformation: %s => %s" % (topology_proposal.old_chemical_state_key, topology_proposal.new_chemical_state_key))
# Determine state keys
old_state_key = self.state_key
new_state_key = topology_proposal.new_chemical_state_key
# Determine log weight
old_log_weight = self.get_log_weight(old_state_key)
new_log_weight = self.get_log_weight(new_state_key)
logp_accept, ncmc_new_sampler_state = self._geometry_ncmc_geometry(topology_proposal, self.sampler.sampler_state, old_log_weight, new_log_weight)
# Accept or reject.
if np.isnan(logp_accept):
accept = False
print('logp_accept = NaN')
else:
accept = ((logp_accept>=0.0) or (np.random.uniform() < np.exp(logp_accept)))
if self.accept_everything:
print('accept_everything option is turned on; accepting')
accept = True
if accept:
self.sampler.thermodynamic_state.set_system(topology_proposal.new_system, fix_state=True)
self.sampler.sampler_state.system = topology_proposal.new_system
self.topology = md.Topology.from_openmm(topology_proposal.new_topology)
self.sampler.sampler_state = ncmc_new_sampler_state
self.sampler.topology = self.topology
self.state_key = topology_proposal.new_chemical_state_key
self.naccepted += 1
if self.verbose: print(" accepted")
else:
self.nrejected += 1
if self.verbose: print(" rejected")
if self.storage:
self.storage.write_configuration('positions', self.sampler.sampler_state.positions, self.topology, iteration=self.iteration)
self.storage.write_object('state_key', self.state_key, iteration=self.iteration)
self.storage.write_object('proposed_state_key', topology_proposal.new_chemical_state_key, iteration=self.iteration)
self.storage.write_quantity('naccepted', self.naccepted, iteration=self.iteration)
self.storage.write_quantity('nrejected', self.nrejected, iteration=self.iteration)
self.storage.write_quantity('logp_accept', logp_accept, iteration=self.iteration)
self.storage.write_quantity('logp_topology_proposal', topology_proposal.logp_proposal, iteration=self.iteration)
# Update statistics.
self.update_statistics()
def update(self):
"""
Update the sampler with one step of sampling.
"""
if self.verbose:
print("-" * 80)
print("Expanded Ensemble sampler iteration %8d" % self.iteration)
self.update_positions(n_iterations=self._n_iterations_per_update)
self.update_state()
self.iteration += 1
if self.verbose:
print("-" * 80)
if self.pdbfile is not None:
print("Writing frame...")
from simtk.openmm.app import PDBFile
PDBFile.writeModel(self.topology.to_openmm(), self.sampler.sampler_state.positions, self.pdbfile, self.iteration)
self.pdbfile.flush()
if self.storage:
self.storage.sync()
def run(self, niterations=1):
"""
Run the sampler for the specified number of iterations
Parameters
----------
niterations : int, optional, default=1
Number of iterations to run the sampler for.
"""
for iteration in range(niterations):
self.update()
def update_statistics(self):
"""
Update sampler statistics.
"""
if self.state_key not in self.number_of_state_visits:
self.number_of_state_visits[self.state_key] = 0
self.number_of_state_visits[self.state_key] += 1
def __init__(self,
temperature=default_temperature,
functions=None,
nsteps=default_nsteps,
steps_per_propagation=default_steps_per_propagation,
timestep=default_timestep,
constraint_tolerance=None,
platform=None,
write_ncmc_interval=1,
measure_shadow_work=False,
integrator_splitting='V R O H R V',
storage=None,
verbose=False,
LRUCapacity=10,
pressure=None,
bond_softening_constant=1.0,
angle_softening_constant=1.0):
"""
This is the base class for NCMC switching between two different systems.
Parameters
----------
temperature : simtk.unit.Quantity with units compatible with kelvin
The temperature at which switching is to be run
functions : dict of str:str, optional, default=default_functions
functions[parameter] is the function (parameterized by 't' which switched from 0 to 1) that
controls how alchemical context parameter 'parameter' is switched
nsteps : int, optional, default=1
The number of steps to use for switching.
steps_per_propagation : int, optional, default=1
The number of intermediate propagation steps taken at each switching step
timestep : simtk.unit.Quantity with units compatible with femtoseconds, optional, default=1*femtosecond
The timestep to use for integration of switching velocity Verlet steps.
constraint_tolerance : float, optional, default=None
If not None, this relative constraint tolerance is used for position and velocity constraints.
platform : simtk.openmm.Platform, optional, default=None
If specified, the platform to use for OpenMM simulations.
write_ncmc_interval : int, optional, default=None
If a positive integer is specified, a snapshot frame will be written to storage with the specified interval on NCMC switching.
'storage' must also be specified.
measure_shadow_work : bool, optional, default False
Whether to measure shadow work
integrator_splitting : str, optional, default='V R O H R V'
NCMC internal integrator splitting based on OpenMMTools Langevin splittings
storage : NetCDFStorageView, optional, default=None
If specified, write data using this class.
verbose : bool, optional, default=False
If True, print debug information.
LRUCapacity : int, default 10
Capacity of LRU cache for hybrid systems
pressure : float, default None
The pressure to use for the simulation. If None, no barostat
"""
# Handle some defaults.
if functions == None:
functions = LambdaProtocol.default_functions
if nsteps == None:
nsteps = default_nsteps
if timestep == None:
timestep = default_timestep
if temperature == None:
temperature = default_temperature
self._temperature = temperature
self._functions = copy.deepcopy(functions)
self._nsteps = nsteps
self._timestep = timestep
self._constraint_tolerance = constraint_tolerance
self._platform = platform
self._integrator_splitting = integrator_splitting
self._steps_per_propagation = steps_per_propagation
self._verbose = verbose
self._pressure = pressure
self._bond_softening_constant = bond_softening_constant
self._angle_softening_constant = angle_softening_constant
self._disable_barostat = False
self._hybrid_cache = LRUCache(capacity=LRUCapacity)
self._measure_shadow_work = measure_shadow_work
self._nattempted = 0
self._storage = None
if storage is not None:
self._storage = NetCDFStorageView(storage,
modname=self.__class__.__name__)
self._save_configuration = True
else:
self._save_configuration = False
if write_ncmc_interval is not None:
self._write_ncmc_interval = write_ncmc_interval
else:
self._write_ncmc_interval = 1
self._work_save_interval = write_ncmc_interval
class NCMCEngine(object):
"""
NCMC switching engine
Examples
--------
Create a transformation for an alanine dipeptide test system where the N-methyl group is eliminated.
>>> from openmmtools import testsystems
>>> testsystem = testsystems.AlanineDipeptideVacuum()
>>> from perses.rjmc.topology_proposal import TopologyProposal
>>> new_to_old_atom_map = { index : index for index in range(testsystem.system.getNumParticles()) if (index > 3) } # all atoms but N-methyl
>>> topology_proposal = TopologyProposal(old_system=testsystem.system, old_topology=testsystem.topology, old_chemical_state_key='AA', new_chemical_state_key='AA', new_system=testsystem.system, new_topology=testsystem.topology, logp_proposal=0.0, new_to_old_atom_map=new_to_old_atom_map, metadata=dict())
>>> ncmc_engine = NCMCEngine(temperature=300.0*unit.kelvin, functions=default_functions, nsteps=50, timestep=1.0*unit.femtoseconds)
>>> positions = testsystem.positions
>>> [positions, logP_delete, potential_delete] = ncmc_engine.integrate(topology_proposal, positions, direction='delete')
>>> [positions, logP_insert, potential_insert] = ncmc_engine.integrate(topology_proposal, positions, direction='insert')
"""
def __init__(self, temperature=default_temperature, functions=None, nsteps=default_nsteps,
steps_per_propagation=default_steps_per_propagation, timestep=default_timestep,
constraint_tolerance=None, platform=None, write_ncmc_interval=1, measure_shadow_work=False,
integrator_splitting='V R O H R V', storage=None, verbose=False, LRUCapacity=10, pressure=None, bond_softening_constant=1.0, angle_softening_constant=1.0):
"""
This is the base class for NCMC switching between two different systems.
Arguments
---------
temperature : simtk.unit.Quantity with units compatible with kelvin
The temperature at which switching is to be run
functions : dict of str:str, optional, default=default_functions
functions[parameter] is the function (parameterized by 't' which switched from 0 to 1) that
controls how alchemical context parameter 'parameter' is switched
nsteps : int, optional, default=1
The number of steps to use for switching.
steps_per_propagation : int, optional, default=1
The number of intermediate propagation steps taken at each switching step
timestep : simtk.unit.Quantity with units compatible with femtoseconds, optional, default=1*femtosecond
The timestep to use for integration of switching velocity Verlet steps.
constraint_tolerance : float, optional, default=None
If not None, this relative constraint tolerance is used for position and velocity constraints.
platform : simtk.openmm.Platform, optional, default=None
If specified, the platform to use for OpenMM simulations.
write_ncmc_interval : int, optional, default=None
If a positive integer is specified, a snapshot frame will be written to storage with the specified interval on NCMC switching.
'storage' must also be specified.
measure_shadow_work : bool, optional, default False
Whether to measure shadow work
integrator_splitting : str, optional, default='V R O H R V'
NCMC internal integrator splitting based on OpenMMTools Langevin splittings
storage : NetCDFStorageView, optional, default=None
If specified, write data using this class.
verbose : bool, optional, default=False
If True, print debug information.
LRUCapacity : int, default 10
Capacity of LRU cache for hybrid systems
pressure : float, default None
The pressure to use for the simulation. If None, no barostat
"""
# Handle some defaults.
if functions == None:
functions = python_hybrid_functions
if nsteps == None:
nsteps = default_nsteps
if timestep == None:
timestep = default_timestep
if temperature == None:
temperature = default_temperature
self._temperature = temperature
self._functions = copy.deepcopy(functions)
self._nsteps = nsteps
self._timestep = timestep
self._constraint_tolerance = constraint_tolerance
self._platform = platform
self._integrator_splitting = integrator_splitting
self._steps_per_propagation = steps_per_propagation
self._verbose = verbose
self._pressure = pressure
self._bond_softening_constant = bond_softening_constant
self._angle_softening_constant = angle_softening_constant
self._disable_barostat = False
self._hybrid_cache = LRUCache(capacity=LRUCapacity)
self._measure_shadow_work = measure_shadow_work
self._nattempted = 0
self._storage = None
if storage is not None:
self._storage = NetCDFStorageView(storage, modname=self.__class__.__name__)
self._save_configuration = True
else:
self._save_configuration = False
if write_ncmc_interval is not None:
self._write_ncmc_interval = write_ncmc_interval
else:
self._write_ncmc_interval = 1
self._work_save_interval = write_ncmc_interval
@property
def beta(self):
kT = kB * self._temperature
beta = 1.0 / kT
return beta
def _compute_energy_contribution(self, hybrid_thermodynamic_state, initial_sampler_state, final_sampler_state):
"""
Compute NCMC energy contribution to log probability.
See Eqs. 62 and 63 (two-stage) and Eq. 45 (hybrid) of reference document.
In both cases, the contribution is u(final_positions, final_lambda) - u(initial_positions, initial_lambda).
Parameters
----------
hybrid_thermodynamic_state : openmmtools.states.CompoundThermodynamicState
The thermodynamic state of the hybrid sampler.
initial_sampler_state : openmmtools.states.SamplerState
The sampler state of the nonalchemical system at the start of the NCMC protocol with box vectors
final_sampler_state : openmmtools.states.SamplerState
The sampler state of the nonalchemical system at the end of the NCMC protocol
Returns
-------
logP_energy : float
The NCMC energy contribution to log probability.
"""
hybrid_thermodynamic_state.set_alchemical_parameters(0.0)
initial_reduced_potential = compute_reduced_potential(hybrid_thermodynamic_state, initial_sampler_state)
hybrid_thermodynamic_state.set_alchemical_parameters(1.0)
final_reduced_potential = compute_reduced_potential(hybrid_thermodynamic_state, final_sampler_state)
return final_reduced_potential - initial_reduced_potential
def _topology_proposal_to_thermodynamic_states(self, topology_proposal):
"""
Convert a topology proposal to thermodynamic states for the end systems. This will be used to compute the
"logP_energy" quantity.
Arguments
---------
topology_proposal : perses.rjmc.TopologyProposal
topology proposal for whose endpoint systems we want ThermodynamicStates
Returns
-------
old_thermodynamic_state : openmmtools.states.ThermodynamicState
The old system (nonalchemical) thermodynamic state
new_thermodynamic_state : openmmtools.states.ThermodynamicState
The new system (nonalchemical) thermodynamic state
"""
systems = [topology_proposal.old_system, topology_proposal.new_system]
thermostates = []
for system in systems:
thermodynamic_state = ThermodynamicState(system, temperature=self._temperature, pressure=self._pressure)
thermostates.append(thermodynamic_state)
return thermostates[0], thermostates[1]
def make_alchemical_system(self, topology_proposal, current_positions, new_positions):
"""
Generate an alchemically-modified system at the correct atoms
based on the topology proposal. This method generates a hybrid system using the new
HybridTopologyFactory. It memoizes so that calling multiple times (within a recent time period)
will immediately return a cached object.
Arguments
---------
topology_proposal : perses.rjmc.TopologyProposal
Unmodified real system corresponding to appropriate leg of transformation.
current_positions : np.ndarray of float
Positions of "old" system
new_positions : np.ndarray of float
Positions of "new" system atoms
Returns
-------
hybrid_factory : perses.annihilation.new_relative.HybridTopologyFactory
a factory object containing the hybrid system
"""
try:
hybrid_factory = self._hybrid_cache[topology_proposal]
#If we've retrieved the factory from the cache, update it to include the relevant positions
hybrid_factory._old_positions = current_positions
hybrid_factory._new_positions = new_positions
hybrid_factory._compute_hybrid_positions()
except KeyError:
try:
hybrid_factory = HybridTopologyFactory(topology_proposal, current_positions, new_positions, bond_softening_constant=self._bond_softening_constant, angle_softening_constant=self._angle_softening_constant)
self._hybrid_cache[topology_proposal] = hybrid_factory
except:
hybrid_factory = None
return hybrid_factory
def integrate(self, topology_proposal, initial_sampler_state, proposed_sampler_state, iteration=None):
"""
Performs NCMC switching to either delete or insert atoms according to the provided `topology_proposal`.
For `delete`, the system is first modified from fully interacting to alchemically modified, and then NCMC switching is used to eliminate atoms.
For `insert`, the system begins with eliminated atoms in an alchemically noninteracting form and NCMC switching is used to turn atoms on, followed by making system real.
Parameters
----------
topology_proposal : TopologyProposal
Contains old/new Topology and System objects and atom mappings.
initial_sampler_state : openmmtools.states.SamplerState representing the initial (old) system
Configurational properties of the atoms at the beginning of the NCMC switching.
proposed_sampler_state : openmmtools.states.SamplerState representing the proposed (post-geometry new) system
Configurational properties new system atoms at beginning of NCMC switching
iteration : int, optional, default=None
Iteration number, for storage purposes.
Returns
-------
final_old_sampler_state : openmmtools.State.SamplerState
The final configurational properties of the old system after hybrid alchemical switching
final_sampler_state : openmmtools.states.SamplerState
The final configurational properties after `nsteps` steps of alchemical switching, and reversion to the nonalchemical system
logP_work : float
The NCMC work contribution to the log acceptance probability (Eqs. 62 and 63)
logP_initial : float
The initial logP of the hybrid configuration
logP_final : float
The final logP of the hybrid configuration
"""
assert not initial_sampler_state.has_nan() and not proposed_sampler_state.has_nan()
#generate or retrieve the hybrid topology factory:
hybrid_factory = self.make_alchemical_system(topology_proposal, initial_sampler_state.positions, proposed_sampler_state.positions)
if hybrid_factory is None:
_logger.warning("Unable to construct hybrid system for {} -> {}".format(topology_proposal.old_chemical_state_key, topology_proposal.new_chemical_state_key))
return initial_sampler_state, proposed_sampler_state, -np.inf, 0.0, 0.0
topology = hybrid_factory.hybrid_topology
#generate the corresponding thermodynamic and sampler states so that we can use the NonequilibriumSwitchingMove:
#First generate the thermodynamic state:
hybrid_system = hybrid_factory.hybrid_system
hybrid_thermodynamic_state = ThermodynamicState(hybrid_system, temperature=self._temperature, pressure=self._pressure)
#Now create an RelativeAlchemicalState from the hybrid system:
alchemical_state = RelativeAlchemicalState.from_system(hybrid_system)
alchemical_state.set_alchemical_parameters(0.0)
#Now create a compound thermodynamic state that combines the hybrid thermodynamic state with the alchemical state:
compound_thermodynamic_state = CompoundThermodynamicState(hybrid_thermodynamic_state, composable_states=[alchemical_state])
#construct a sampler state from the hybrid positions and the box vectors of the initial sampler state:
initial_hybrid_positions = hybrid_factory.hybrid_positions
initial_hybrid_box_vectors = initial_sampler_state.box_vectors
initial_hybrid_sampler_state = SamplerState(initial_hybrid_positions, box_vectors=initial_hybrid_box_vectors)
final_hybrid_sampler_state = copy.deepcopy(initial_hybrid_sampler_state)
#create the nonequilibrium move:
#ne_move = NonequilibriumSwitchingMove(self._functions, self._integrator_splitting, self._temperature, self._nsteps, self._timestep,
# work_save_interval=self._write_ncmc_interval, top=topology,subset_atoms=None,
# save_configuration=self._save_configuration, measure_shadow_work=self._measure_shadow_work)
ne_move = ExternalNonequilibriumSwitchingMove(self._functions, nsteps_neq=self._nsteps,
timestep=self._timestep, temperature=self._temperature,
work_configuration_save_interval=self._work_save_interval,
splitting="V R O R V")
#run the NCMC protocol
try:
ne_move.apply(compound_thermodynamic_state, final_hybrid_sampler_state)
except Exception as e:
_logger.warn("NCMC failed because {}; rejecting.".format(str(e)))
logP_work = -np.inf
return [initial_sampler_state, proposed_sampler_state, -np.inf, 0.0, 0.0]
#get the total work:
logP_work = - ne_move.cumulative_work[-1]
# Compute contribution of transforming to and from the hybrid system:
context, integrator = global_context_cache.get_context(hybrid_thermodynamic_state)
#set all alchemical parameters to zero:
for parameter in self._functions.keys():
context.setParameter(parameter, 0.0)
initial_hybrid_sampler_state.apply_to_context(context, ignore_velocities=True)
initial_reduced_potential = hybrid_thermodynamic_state.reduced_potential(context)
#set all alchemical parameters to one:
for parameter in self._functions.keys():
context.setParameter(parameter, 1.0)
final_hybrid_sampler_state.apply_to_context(context, ignore_velocities=True)
final_reduced_potential = hybrid_thermodynamic_state.reduced_potential(context)
#reset the parameters back to zero just in case
for parameter in self._functions.keys():
context.setParameter(parameter, 0.0)
#compute the output SamplerState, which has the atoms only for the new system post-NCMC:
new_positions = hybrid_factory.new_positions(final_hybrid_sampler_state.positions)
new_box_vectors = final_hybrid_sampler_state.box_vectors
final_sampler_state = SamplerState(new_positions, box_vectors=new_box_vectors)
#compute the output SamplerState for the atoms only in the old system (required for geometry_logP_reverse)
old_positions = hybrid_factory.old_positions(final_hybrid_sampler_state.positions)
old_box_vectors = copy.deepcopy(new_box_vectors) #these are the same as the new system
final_old_sampler_state = SamplerState(old_positions, box_vectors=old_box_vectors)
#extract the trajectory and box vectors from the move:
trajectory = ne_move.trajectory[::-self._write_ncmc_interval, :, :][::-1]
topology = hybrid_factory.hybrid_topology
position_varname = "ncmcpositions"
nframes = np.shape(trajectory)[0]
#extract box vectors:
box_vec_varname = "ncmcboxvectors"
box_lengths = ne_move.box_lengths[::-self._write_ncmc_interval, :][::-1]
box_angles = ne_move.box_angles[::-self._write_ncmc_interval, :][::-1]
box_lengths_and_angles = np.stack([box_lengths, box_angles])
#write out the positions of the topology
if self._storage:
for frame in range(nframes):
self._storage.write_configuration(position_varname, trajectory[frame, :, :], topology, iteration=iteration, frame=frame, nframes=nframes)
#write out the periodict box vectors:
if self._storage:
self._storage.write_array(box_vec_varname, box_lengths_and_angles, iteration=iteration)
#retrieve the protocol work and write that out too:
protocol_work = ne_move.cumulative_work
if self._storage:
self._storage.write_array("protocolwork", protocol_work, iteration=iteration)
# Return
return [final_old_sampler_state, final_sampler_state, logP_work, -initial_reduced_potential, -final_reduced_potential]
