def run_selectivity():
storage_filename = 'output-src-abl.nc'
system = Selectivity(storage_filename=storage_filename)
system.designer.run(niterations=10)
from perses.analysis import Analysis
analysis = Analysis(storage_filename=storage_filename)
analysis.plot_ncmc_work('ncmc-work.pdf')
def run_selectivity():
storage_filename = 'output-src-abl.nc'
system = Selectivity(storage_filename=storage_filename)
system.designer.run(niterations=10)
from perses.analysis import Analysis
analysis = Analysis(storage_filename=storage_filename)
analysis.plot_ncmc_work('ncmc-work.pdf')
def collect_logP_accept(condition_files):
"""
Given a set of files specifying conditions, extract the logP_accept of each and store in a data structure
Parameters
----------
condition_files : dict of tuple: str
Should have the format (phase, ncmc_length, sterics, geometry_intervals) : filename
Returns
-------
logP_accept_conditions: dict of tuple: np.array
the logP_accept (minus sams weights) for each set of conditions
"""
logP_accept_conditions = {}
for condition, filename in condition_files.items():
try:
analyzer = Analysis(filename)
logP_with_sams = analyzer._ncfile.groups[condition[0]]['ExpandedEnsembleSampler']['logP_accept'][:]
sams = analyzer._ncfile.groups[condition[0]]['ExpandedEnsembleSampler']['logP_sams'][:]
logP_without_sams = logP_with_sams - sams
logP_unmasked = logP_without_sams[~logP_without_sams.mask]
logP_accept_conditions[condition] = logP_unmasked
except Exception as e:
print(str(e))
print("Unable to process {}".format(filename))
continue
return logP_accept_conditions
def test_analysis():
"""Test analysis tools.
"""
testsystem_names = ['AlanineDipeptideTestSystem']
for testsystem_name in testsystem_names:
# Create storage.
tmpfile = tempfile.NamedTemporaryFile()
storage_filename = tmpfile.name
import perses.tests.testsystems
testsystem_class = getattr(perses.tests.testsystems, testsystem_name)
# Instantiate test system.
testsystem = testsystem_class(storage_filename=storage_filename)
# Alter settings
for environment in testsystem.environments:
testsystem.mcmc_samplers[environment].verbose = False
testsystem.mcmc_samplers[
environment].nsteps = 5 # use fewer MD steps to speed things up
testsystem.exen_samplers[environment].verbose = False
testsystem.exen_samplers[
environment].ncmc_engine.nsteps = 5 # NCMC switching
testsystem.sams_samplers[environment].verbose = False
# Run test simulations.
niterations = 5 # just a few iterations
if testsystem.designer is not None:
# Run the designer
testsystem.designer.verbose = False
testsystem.designer.run(niterations=niterations)
else:
# Run the SAMS samplers.
for environment in testsystem.environments:
testsystem.sams_samplers[environment].run(
niterations=niterations)
# Analyze file.
# TODO: Use temporary filenames
analysis = Analysis(storage_filename)
analysis.plot_ncmc_work('ncmc.pdf')
def test_analysis():
"""Test analysis tools.
"""
testsystem_names = ['AlanineDipeptideTestSystem']
for testsystem_name in testsystem_names:
# Create storage.
tmpfile = tempfile.NamedTemporaryFile()
storage_filename = tmpfile.name
import perses.tests.testsystems
testsystem_class = getattr(perses.tests.testsystems, testsystem_name)
# Instantiate test system.
testsystem = testsystem_class(storage_filename=storage_filename)
# Alter settings
for environment in testsystem.environments:
testsystem.mcmc_samplers[environment].verbose = False
testsystem.mcmc_samplers[environment].nsteps = 5 # use fewer MD steps to speed things up
testsystem.exen_samplers[environment].verbose = False
testsystem.exen_samplers[environment].ncmc_engine.nsteps = 5 # NCMC switching
testsystem.sams_samplers[environment].verbose = False
# Run test simulations.
niterations = 5 # just a few iterations
if testsystem.designer is not None:
# Run the designer
testsystem.designer.verbose = False
testsystem.designer.run(niterations=niterations)
else:
# Run the SAMS samplers.
for environment in testsystem.environments:
testsystem.sams_samplers[environment].run(niterations=niterations)
# Analyze file.
# TODO: Use temporary filenames
analysis = Analysis(storage_filename)
analysis.plot_ncmc_work('ncmc.pdf')
def benchmark_ncmc_work_during_protocol():
"""
Run 50 iterations of ExpandedEnsembleSampler for NullTestSystems
over a range of total NCMC steps [0, 1, 10, 100, 1000, 10000].
Benchmark is repeated for Naphthalene, Butane, and Propane test
systems, using two-stage and hybrid NCMC.
For each combination of system and scheme, results are analyzed for
the following:
* For a given total number of steps:
* For NCMC steps 100 and above, plot work done by ncmc integrator
over the course of the protocol
* Plot histograms of the contributions of each component to the
overall log acceptance probability
* Over the whole range of total steps:
* Plot mean and standard deviation of NCMC logP as a function of
total steps
* Plot mean and standard deviation of EXEN logP as a function of
total steps
"""
from perses.tests.testsystems import NaphthaleneTestSystem, ButaneTestSystem, PropaneTestSystem
from perses.analysis import Analysis
import netCDF4 as netcdf
import pickle
import codecs
molecule_names = {
#'propane' : PropaneTestSystem,
'butane': ButaneTestSystem,
#'naphthalene' : NaphthaleneTestSystem,
}
methods = {
'hybrid': ['geometry-ncmc-geometry', functions_hybrid],
'two-stage': ['ncmc-geometry-ncmc', functions_twostage],
}
for molecule_name, NullProposal in molecule_names.items():
print('\nNow testing {0} null transformations'.format(molecule_name))
for name, [scheme, functions] in methods.items():
analyses = dict()
#for ncmc_nsteps in [0, 1, 10, 100, 1000, 10000]:
for ncmc_nsteps in [0, 1, 10, 100, 1000]:
print(
'Running {0} {2} ExpandedEnsemble steps for {1} iterations'
.format(ncmc_nsteps, niterations, name))
testsystem = NullProposal(
storage_filename='{0}_{1}-{2}steps.nc'.format(
molecule_name, name, ncmc_nsteps),
scheme=scheme,
options={
'functions': functions,
'nsteps': ncmc_nsteps
})
testsystem.exen_samplers[
ENV].geometry_engine.use_sterics = use_sterics
testsystem.mcmc_samplers[ENV].verbose = True
testsystem.exen_samplers[ENV].verbose = True
testsystem.mcmc_samplers[
ENV].timestep = 1.0 * unit.femtoseconds
# DEBUG
#testsystem.exen_samplers[ENV].geometry_engine.write_proposal_pdb = True
#testsystem.exen_samplers[ENV].geometry_engine.pdb_filename_prefix = '{0}_{1}-{2}steps'.format(molecule_name, name, ncmc_nsteps)
# Equilibrate
# WARNING: We can't equilibrate because it messes up the iteration counters and storage iterations for exen samplers
#print('Equilibration...')
#testsystem.mcmc_samplers[ENV].run(niterations=nequil)
# Collect data on switching
print('Production...')
testsystem.exen_samplers[ENV].run(niterations=niterations)
analysis = Analysis(testsystem.storage_filename)
print(analysis.get_environments())
if ncmc_nsteps > 9:
analysis.plot_ncmc_work(
'{0}_{1}-ncmc_work_over_{2}_steps.pdf'.format(
molecule_name, name, ncmc_nsteps))
analysis.plot_exen_logp_components()
analyses[ncmc_nsteps] = analysis
benchmark_exen_ncmc_protocol(analyses, molecule_name, name)
def benchmark_ncmc_work_during_protocol():
"""
Run 50 iterations of ExpandedEnsembleSampler for NullTestSystems
over a range of total NCMC steps [0, 1, 10, 100, 1000, 10000].
Benchmark is repeated for Naphthalene, Butane, and Propane test
systems, using two-stage and hybrid NCMC.
For each combination of system and scheme, results are analyzed for
the following:
* For a given total number of steps:
* For NCMC steps 100 and above, plot work done by ncmc integrator
over the course of the protocol
* Plot histograms of the contributions of each component to the
overall log acceptance probability
* Over the whole range of total steps:
* Plot mean and standard deviation of NCMC logP as a function of
total steps
* Plot mean and standard deviation of EXEN logP as a function of
total steps
"""
from perses.tests.testsystems import NaphthaleneTestSystem, ButaneTestSystem, PropaneTestSystem
from perses.analysis import Analysis
import netCDF4 as netcdf
import pickle
import codecs
molecule_names = {
#'propane' : PropaneTestSystem,
'butane' : ButaneTestSystem,
#'naphthalene' : NaphthaleneTestSystem,
}
methods = {
'hybrid' : ['geometry-ncmc-geometry', functions_hybrid],
'two-stage' : ['ncmc-geometry-ncmc', functions_twostage],
}
for molecule_name, NullProposal in molecule_names.items():
print('\nNow testing {0} null transformations'.format(molecule_name))
for name, [scheme, functions] in methods.items():
analyses = dict()
#for ncmc_nsteps in [0, 1, 10, 100, 1000, 10000]:
for ncmc_nsteps in [0, 1, 10, 100, 1000]:
print('Running {0} {2} ExpandedEnsemble steps for {1} iterations'.format(ncmc_nsteps, niterations, name))
testsystem = NullProposal(storage_filename='{0}_{1}-{2}steps.nc'.format(molecule_name, name, ncmc_nsteps), scheme=scheme, options={'functions' : functions, 'nsteps' : ncmc_nsteps})
testsystem.exen_samplers[ENV].geometry_engine.use_sterics = use_sterics
testsystem.mcmc_samplers[ENV].verbose = True
testsystem.exen_samplers[ENV].verbose = True
testsystem.mcmc_samplers[ENV].timestep = 1.0 * unit.femtoseconds
# DEBUG
#testsystem.exen_samplers[ENV].geometry_engine.write_proposal_pdb = True
#testsystem.exen_samplers[ENV].geometry_engine.pdb_filename_prefix = '{0}_{1}-{2}steps'.format(molecule_name, name, ncmc_nsteps)
# Equilibrate
# WARNING: We can't equilibrate because it messes up the iteration counters and storage iterations for exen samplers
#print('Equilibration...')
#testsystem.mcmc_samplers[ENV].run(niterations=nequil)
# Collect data on switching
print('Production...')
testsystem.exen_samplers[ENV].run(niterations=niterations)
analysis = Analysis(testsystem.storage_filename)
print(analysis.get_environments())
if ncmc_nsteps > 9:
analysis.plot_ncmc_work('{0}_{1}-ncmc_work_over_{2}_steps.pdf'.format(molecule_name, name, ncmc_nsteps))
analysis.plot_exen_logp_components()
analyses[ncmc_nsteps] = analysis
benchmark_exen_ncmc_protocol(analyses, molecule_name, name)