def test_against_vmd(pdb, get_fn):
pdb = get_fn(pdb)
# this is probably not cross-platform compatible. I assume that the exact
# path to this CHARMM topology that is included with VMD depends on
# the install mechanism, especially for bundled mac or windows installers
VMD_ROOT = os.path.join(os.path.dirname(os.path.realpath(VMD)), '..')
top_paths = [os.path.join(r, f) for (r, _, fs) in os.walk(VMD_ROOT) for f in fs
if 'top_all27_prot_lipid_na.inp' in f]
assert len(top_paths) >= 0
top = os.path.abspath(top_paths[0]).replace(" ", "\\ ")
TEMPLATE = '''
package require psfgen
topology %(top)s
pdbalias residue HIS HSE
pdbalias atom ILE CD1 CD
segment U {pdb %(pdb)s}
coordpdb %(pdb)s U
guesscoord
writepdb out.pdb
writepsf out.psf
exit
''' % {'top': top, 'pdb': pdb}
with enter_temp_directory():
with open('script.tcl', 'w') as f:
f.write(TEMPLATE)
subprocess.check_call([VMD, '-startup', 'script.tcl', '-dispdev', 'none'])
out_pdb = md.load('out.pdb')
out_psf = md.load_psf('out.psf')
# make sure the two topologies are equal
eq(out_pdb.top, out_psf)
def _test_against_vmd(pdb):
# this is probably not cross-platform compatible. I assume that the exact
# path to this CHARMM topology that is included with VMD depends on
# the install mechanism, especially for bundled mac or windows installers
VMD_ROOT = os.path.join(os.path.dirname(os.path.realpath(VMD)), '..')
top_paths = [os.path.join(r, f) for (r, _, fs) in os.walk(VMD_ROOT) for f in fs
if 'top_all27_prot_lipid_na.inp' in f]
assert len(top_paths) >= 0
top = os.path.abspath(top_paths[0]).replace(" ", "\\ ")
TEMPLATE = '''
package require psfgen
topology %(top)s
pdbalias residue HIS HSE
pdbalias atom ILE CD1 CD
segment U {pdb %(pdb)s}
coordpdb %(pdb)s U
guesscoord
writepdb out.pdb
writepsf out.psf
exit
''' % {'top': top, 'pdb' : pdb}
with enter_temp_directory():
with open('script.tcl', 'w') as f:
f.write(TEMPLATE)
os.system(' '.join([VMD, '-e', 'script.tcl', '-dispdev', 'none']))
out_pdb = md.load('out.pdb')
out_psf = md.load_psf('out.psf')
# make sure the two topologies are equal
eq(out_pdb.top, out_psf)
def __init__(self, true_value=None, initial_value=None, n_increments=18, rj=True, sample_phase=False,
continuous=False):
self._param = CharmmParameterSet(get_fun('toy.str'))
self._struct = CharmmPsfFile(get_fun('toy.psf'))
self._pdb = app.PDBFile(get_fun('toy.pdb'))
self._topology = md.load_psf(get_fun('toy.psf'))
self.synthetic_energy = units.Quantity()
self._positions = units.Quantity()
self._platform = mm.Platform.getPlatformByName('Reference')
# Replace ('CG331', 'CG321', 'CG321', 'CG331') torsion with true_value
self._dih_type = ('CG331', 'CG321', 'CG321', 'CG331')
original_torsion = self._param.dihedral_types[self._dih_type]
if true_value is not None:
if type(true_value) == DihedralTypeList:
dih_tlist = true_value
elif type(true_value) == DihedralType:
dih_tlist = DihedralTypeList()
dih_tlist.append(true_value)
else:
dih_tlist = self._randomize_dih_param(return_dih=True)
self.true_value = copy.deepcopy(dih_tlist)
self._param.dihedral_types[self._dih_type] = dih_tlist
# parametrize toy
self._struct.load_parameters(self._param, copy_parameters=False)
self._struct.positions = self._pdb.positions
# generate synthetic torsion scan
self._torsion_scan(n_increments=n_increments)
# initialize parameter
if initial_value is not None:
if type(initial_value) == DihedralTypeList:
dih_tlist = initial_value
if type(initial_value) == DihedralType:
dih_tlist = DihedralTypeList()
dih_tlist.append(initial_value)
elif initial_value == 'cgenff':
dih_tlist = original_torsion
else:
dih_tlist = self._randomize_dih_param(return_dih=True)
self.initial_value = copy.deepcopy(dih_tlist)
self._param.dihedral_types[self._dih_type] = dih_tlist
# create torsionfit.TorsionScanSet
torsions = np.zeros((len(self._positions), 4))
torsions[:] = [1, 2, 3, 4]
direction = None
steps = None
self.scan_set = ScanSet.QMDataBase(positions=self._positions.value_in_unit(units.nanometers),
topology=self._topology, structure=self._struct, torsions=torsions,
steps=steps, directions=direction,
qm_energies=self.synthetic_energy.value_in_unit(units.kilojoules_per_mole))
self.model = model.TorsionFitModel(param=self._param, frags=self.scan_set, platform=self._platform,
param_to_opt=[self._dih_type], rj=rj, continuous_phase=continuous,
sample_phase=sample_phase)
def read_scan_logfile(logfiles, structure):
""" parses Guassian09 torsion-scan log file
parameters
----------
logfiles: str of list of str
Name of Guassian 09 torsion scan log file
structure: charmm psf file
returns
-------
TorsionScanSet
"""
topology = md.load_psf(structure)
structure = CharmmPsfFile(structure)
positions = np.ndarray((0, topology.n_atoms, 3))
qm_energies = np.ndarray(0)
torsions = np.ndarray((0, 4), dtype=int)
directions = np.ndarray(0, dtype=int)
steps = np.ndarray((0, 3), dtype=int)
if type(logfiles) != list:
logfiles = [logfiles]
for file in logfiles:
print("loading %s" % file)
direction = np.ndarray(1)
torsion = np.ndarray((1, 4), dtype=int)
step = []
index = (2, 12, -1)
f = file.split("/")[-1].split(".")
if f[2] == "pos":
direction[0] = 1
else:
direction[0] = 0
fi = open(file, "r")
for line in fi:
if re.search(" Scan ", line):
t = line.split()[2].split(",")
t[0] = t[0][-1]
t[-1] = t[-1][0]
for i in range(len(t)):
torsion[0][i] = int(t[i]) - 1
if re.search("Step", line):
try:
step = np.array(([int(line.rsplit()[j]) for j in index]))
step = step[np.newaxis, :]
steps = np.append(steps, step, axis=0)
except:
pass
fi.close()
log = Gaussian(file)
data = log.parse()
# convert angstroms to nanometers
positions = np.append(positions, data.atomcoords * 0.1, axis=0)
qm_energies = np.append(
qm_energies,
(convertor(data.scfenergies, "eV", "kJmol-1") - min(convertor(data.scfenergies, "eV", "kJmol-1"))),
axis=0,
)
for i in range(len(data.scfenergies)):
torsions = np.append(torsions, torsion, axis=0)
directions = np.append(directions, direction, axis=0)
return TorsionScanSet(positions, topology, structure, torsions, directions, steps, qm_energies)
def parse_psi4_log(logfiles, structure):
"""
Parses output of psi4 torsion scan script
:param logfiles: list of str
logfiles of psi4 script
:param structure: str
Charmm psf file of structure
:return:
TorsionScanSet
"""
topology = md.load_psf(structure)
structure = CharmmPsfFile(structure)
positions = np.ndarray((0, topology.n_atoms, 3))
qm_energies = np.ndarray(0)
torsions = np.ndarray((0, 4), dtype=int)
directions = np.ndarray(0, dtype=int)
angles = np.ndarray(0, dtype=float)
if type(logfiles) != list:
logfiles = [logfiles]
for file in logfiles:
qm = np.ndarray(0)
fi = open(file, 'r')
# check if log file is complete
complete = False # complete flag
for line in fi:
if line.startswith('Relative'):
complete = True
fi.seek(0)
section = None
torsion = np.ndarray((1, 4), dtype=int)
angle = np.ndarray(0, dtype=float)
for line in fi:
# Flag if structure is optimized
optimized = False
if line.startswith('Optimizer'):
optimized = True
# Store Dihedral and position of optimized structures
fi.next()
l = filter(None, fi.next().strip().split(' '))
dih = round(float(l[-2]))
try:
t = l[-6:-2]
for i in range(len(t)):
torsion[0][i] = int(t[i]) - 1
torsions = np.append(torsions, torsion, axis=0)
except ValueError:
pass
angle = np.append(angle, dih)
fi.next()
pos = filter(None, re.split("[, \[\]]", fi.next().strip()))
pos = [float(i) for i in pos]
pos = np.asarray(pos).reshape((-1, 3))
# convert angstroms to nanometers
positions = np.append(positions, pos[np.newaxis] * 0.1, axis=0)
if not complete and optimized:
# Find line that starts with energy
for line in fi:
if line.startswith('Energy'):
energy = filter(None, line.strip().split(' '))[-1]
# Convert to KJ/mol
energy = float(energy) * 2625.5
qm = np.append(qm, energy)
break
if line.startswith('Relative'):
section = 'Energy'
fi.next()
continue
if section == 'Energy':
line = filter(None, line.strip().split(' '))
if line != []:
dih = round(float(line[0]))
if dih in angle:
# Only save energies of optimized structures
qm_energies = np.append(qm_energies, float(line[-1]))
if qm.size is not 0:
qm = qm - min(qm)
qm_energies = np.append(qm_energies, qm)
fi.close()
angles = np.append(angles, angle, axis=0)
return QMDataBase(positions, topology, structure, torsions, directions,
angles, qm_energies)
def parse_gauss(logfiles, structure):
""" parses Guassian09 torsion-scan log file
parameters
----------
logfiles: str of list of str
Name of Guassian 09 torsion scan log file
structure: charmm psf file
returns
-------
TorsionScanSet
"""
topology = md.load_psf(structure)
structure = CharmmPsfFile(structure)
positions = np.ndarray((0, topology.n_atoms, 3))
qm_energies = np.ndarray(0)
torsions = np.ndarray((0, 4), dtype=int)
directions = np.ndarray(0, dtype=int)
steps = np.ndarray((0, 3), dtype=int)
if type(logfiles) != list:
logfiles = [logfiles]
for file in (logfiles):
direction = np.ndarray(1)
torsion = np.ndarray((1, 4), dtype=int)
step = np.ndarray((0, 3), dtype=int)
index = (2, 12, -1)
log = Gaussian(file)
data = log.parse()
# convert angstroms to nanometers
positions = np.append(positions, data.atomcoords * 0.1, axis=0)
# Only add qm energies for structures that converged (because cclib throws out those coords but not other info)
qm_energies = np.append(qm_energies, (convertor(
data.scfenergies[:len(data.atomcoords)], "eV", "kJmol-1") - min(
convertor(data.scfenergies[:len(data.atomcoords)], "eV",
"kJmol-1"))),
axis=0)
fi = open(file, 'r')
for line in fi:
if re.search(' Scan ', line):
t = line.split()[2].split(',')
t[0] = t[0][-1]
t[-1] = t[-1][0]
for i in range(len(t)):
torsion[0][i] = (int(t[i]) - 1)
if re.search('^ D ', line):
d = line.split()[-1]
if d[0] == '-':
direction[0] = 0
elif d[0] == '1':
direction[0] = 1
if re.search('Step', line):
try:
point = np.array(([int(line.rsplit()[j]) for j in index]))
point = point[np.newaxis, :]
step = np.append(step, point, axis=0)
except:
pass
fi.close()
# only add scan points from converged structures
steps = np.append(steps, step[:len(data.atomcoords)], axis=0)
for i in range(len(data.atomcoords)):
torsions = np.append(torsions, torsion, axis=0)
directions = np.append(directions, direction, axis=0)
del log
del data
return QMDataBase(positions=positions,
topology=topology,
structure=structure,
torsions=torsions,
steps=steps,
qm_energies=qm_energies,
directions=directions)
def parse_psi4_out(oufiles_dir, structure, pattern="*.out"):
"""
Parse psi4 out files from distributed torsion scan (there are many output files, one for each structure)
:param oufiles_dir: str
path to directory where the psi4 output files are
:param structure: str
path to psf, mol2 or pbd file of structure
:param pattern: str
pattern for psi4 output file. Default is *.out
:return: TorsionScanSet
"""
# Check extension of structure file
if structure.endswith('psf'):
topology = md.load_psf(structure)
structure = CharmmPsfFile(structure)
else:
topology = md.load(structure).topology
structure = parmed.load_file(structure)
positions = np.ndarray((0, topology.n_atoms, 3))
qm_energies = np.ndarray(0)
torsions = np.ndarray((0, 4), dtype=int)
angles = np.ndarray(0, dtype=float)
optimized = np.ndarray(0, dtype=bool)
out_files = {}
for path, subdir, files in os.walk(oufiles_dir):
for name in files:
if fnmatch(name, pattern):
if name.startswith('timer'):
continue
name_split = name.split('_')
try:
torsion_angle = (name_split[1] + '_' + name_split[2] +
'_' + name_split[3] + '_' + name_split[4])
except IndexError:
warnings.warn(
"Do you only have one torsion scan? The output files will be treated as one scan"
)
torsion_angle = 'only_one_scan'
try:
out_files[torsion_angle]
except KeyError:
out_files[torsion_angle] = []
path = os.path.join(os.getcwd(), path, name)
out_files[torsion_angle].append(path)
# Sort files in increasing angles order for each torsion
sorted_files = []
dih_angles = []
for tor in out_files:
dih_angle = []
for file in out_files[tor]:
dih_angle.append(int(file.split('_')[-1].split('.')[0]))
sorted_files.append([
out_file
for (angle, out_file) in sorted(zip(dih_angle, out_files[tor]))
])
dih_angle.sort()
dih_angles.append(dih_angle)
if not out_files:
raise Exception(
"There are no psi4 output files. Did you choose the right directory?"
)
# Parse files
for f in itertools.chain.from_iterable(sorted_files):
torsion = np.ndarray((1, 4), dtype=int)
fi = open(f, 'r')
for line in fi:
if line.startswith('dih_string'):
t = line.strip().split('"')[1].split(' ')[:4]
for i in range(len(t)):
torsion[0][i] = int(t[i]) - 1
torsions = np.append(torsions, torsion, axis=0)
fi.close()
optimizer = True
log = Psi(f)
data = log.parse()
try:
data.optdone
except AttributeError:
optimizer = False
warnings.warn("Warning: Optimizer failed for {}".format(f))
optimized = np.append(optimized, optimizer)
positions = np.append(positions,
data.atomcoords[-1][np.newaxis] * 0.1,
axis=0)
# Try MP2 energies. Otherwise take SCFenergies
try:
qm_energy = convertor(data.mpenergies[-1], "eV", "kJmol-1")
except AttributeError:
try:
qm_energy = convertor(np.array([data.scfenergies[-1]]), "eV",
"kJmol-1")
except AttributeError:
warnings.warn(
"Warning: Check if the file terminated before completing SCF"
)
qm_energy = np.array([np.nan])
qm_energies = np.append(qm_energies, qm_energy, axis=0)
# Subtract lowest energy to find relative energies
qm_energies = qm_energies - min(qm_energies)
angles = np.asarray(list(itertools.chain.from_iterable(dih_angles)))
return QMDataBase(positions=positions,
topology=topology,
structure=structure,
torsions=torsions,
angles=angles,
qm_energies=qm_energies,
optimized=optimized)
"""
Based on trajectories from umbrella sampling, we compute the butane dihedral and
save them in .csv files.
"""
import numpy as np
import mdtraj
import math
from sys import exit
topology = mdtraj.load_psf("./data/butane.psf")
M = 30
for theta0_index in range(M):
print(theta0_index)
traj = mdtraj.load_dcd(f"./output/traj/traj_{theta0_index}.dcd", topology)
theta = mdtraj.compute_dihedrals(traj, [[3, 6, 9, 13]])
np.savetxt(f"./output/dihedral/dihedral_{theta0_index}.csv",
theta,
fmt="%.5f",
delimiter=",")
__author__ = "Xinqiang Ding <*****@*****.**>"
__date__ = "2019/10/05 02:25:36"
import numpy as np
import matplotlib.pyplot as plt
import mdtraj
import math
import simtk.unit as unit
import sys
#sys.path.insert(0, "/home/xqding/course/projectsOnGitHub/FastMBAR/FastMBAR/")
from FastMBAR import *
from sys import exit
import pickle
topology = mdtraj.load_psf("./output/dialanine.psf")
K = 100
m = 25
M = m * m
psi = np.linspace(-math.pi, math.pi, m, endpoint=False)
phi = np.linspace(-math.pi, math.pi, m, endpoint=False)
psis = []
phis = []
for psi_index in range(m):
for phi_index in range(m):
traj = mdtraj.load_dcd(
f"./output/traj/traj_psi_{psi_index}_phi_{phi_index}.dcd",
topology)
psis.append(mdtraj.compute_dihedrals(traj, [[4, 6, 8, 14]]))
phis.append(mdtraj.compute_dihedrals(traj, [[6, 8, 14, 16]]))
def read_scan_logfile(logfiles, structure):
""" parses Guassian09 torsion-scan log file
parameters
----------
logfiles: str of list of str
Name of Guassian 09 torsion scan log file
structure: charmm psf file
returns
-------
TorsionScanSet
"""
topology = md.load_psf(structure)
structure = CharmmPsfFile(structure)
positions = np.ndarray((0, topology.n_atoms, 3))
qm_energies = np.ndarray(0)
torsions = np.ndarray((0, 4), dtype=int)
directions = np.ndarray(0, dtype=int)
steps = np.ndarray((0, 3), dtype=int)
if type(logfiles) != list:
logfiles = [logfiles]
for file in (logfiles):
#print("loading %s" % file)
direction = np.ndarray(1)
torsion = np.ndarray((1, 4), dtype=int)
step = np.ndarray((0, 3), dtype=int)
index = (2, 12, -1)
# f = file.split('/')[-1].split('.')
# if f[2] == 'pos':
# direction[0] = 1
# else:
# direction[0] = 0
log = Gaussian(file)
data = log.parse()
# convert angstroms to nanometers
positions = np.append(positions, data.atomcoords*0.1, axis=0)
# Only add qm energies for structures that converged (because cclib throws out those coords but not other info)
qm_energies = np.append(qm_energies, (convertor(data.scfenergies[:len(data.atomcoords)], "eV", "kJmol-1") -
min(convertor(data.scfenergies[:len(data.atomcoords)], "eV", "kJmol-1"))), axis=0)
fi = open(file, 'r')
for line in fi:
if re.search(' Scan ', line):
t = line.split()[2].split(',')
t[0] = t[0][-1]
t[-1] = t[-1][0]
for i in range(len(t)):
torsion[0][i] = (int(t[i]) - 1)
if re.search('^ D ', line):
d = line.split()[-1]
if d[0] == '-':
direction[0] = 0
elif d[0] == '1':
direction[0] = 1
if re.search('Step', line):
try:
point = np.array(([int(line.rsplit()[j]) for j in index]))
point = point[np.newaxis,:]
step = np.append(step, point, axis=0)
except:
pass
fi.close()
# only add scan points from converged structures
steps = np.append(steps, step[:len(data.atomcoords)], axis=0)
for i in range(len(data.atomcoords)):
torsions = np.append(torsions, torsion, axis=0)
directions = np.append(directions, direction, axis=0)
del log
del data
return TorsionScanSet(positions, topology, structure, torsions, directions, steps, qm_energies)