def _analyse(self, mol, rtf, prm, pdb, traj):
m = FFMolecule( filename=mol, rtf=rtf, prm=prm)
m.read(traj)
torsions = m.getSortTorsions()
print(torsions)
for i in range(len(torsions[0])):
# For each torsion, measure
title = torsions[1][i][0]
title = title + "-" + torsions[1][i][1]
title = title + "-" + torsions[1][i][2]
title = title + "-" + torsions[1][i][3]
(r, theta) = self._measure_torsion(torsions[0][i], m.coords)
self._plot_scatter(r, theta, title)
self._plot_hist(theta, title)
def test_makeDihedralUnique(self):
from htmd.home import home
from htmd.parameterization.ffmolecule import FFMolecule, FFTypeMethod
molFile = os.path.join(home('test-param'), 'glycol.mol2')
self.df.molecule = FFMolecule(molFile, method=FFTypeMethod.GAFF2)
types = [
self.df.molecule._rtf.type_by_index[i]
for i in range(self.df.molecule.numAtoms)
]
self.assertListEqual(
types,
['oh', 'c3', 'c3', 'oh', 'ho', 'h1', 'h1', 'h1', 'h1', 'ho'])
self.df.molecule = FFMolecule(molFile, method=FFTypeMethod.GAFF2)
self.df._makeDihedralUnique([0, 1, 2, 3])
types = [
self.df.molecule._rtf.type_by_index[i]
for i in range(self.df.molecule.numAtoms)
]
self.assertListEqual(types, [
'ohx0', 'c3x0', 'c3x0', 'ohx0', 'ho', 'h1', 'h1', 'h1', 'h1', 'ho'
])
self.df.molecule = FFMolecule(molFile, method=FFTypeMethod.GAFF2)
self.df._makeDihedralUnique([4, 0, 1, 2])
types = [
self.df.molecule._rtf.type_by_index[i]
for i in range(self.df.molecule.numAtoms)
]
self.assertListEqual(types, [
'ohx0', 'c3x0', 'c3x0', 'ohx0', 'hox0', 'h1', 'h1', 'h1', 'h1',
'hox0'
])
self.df.molecule = FFMolecule(molFile, method=FFTypeMethod.GAFF2)
self.df._makeDihedralUnique([5, 1, 2, 7])
types = [
self.df.molecule._rtf.type_by_index[i]
for i in range(self.df.molecule.numAtoms)
]
self.assertListEqual(types, [
'oh', 'c3x0', 'c3x0', 'oh', 'ho', 'h1x0', 'h1x0', 'h1x0', 'h1x0',
'ho'
])
def main_scan():
if len(sys.argv) != 2:
print("Syntax: %s input.mol" % (sys.argv[0]))
sys.exit(0)
filename = sys.argv[1]
print("Scan {}".format(filename))
mol = FFMolecule(filename=filename, method=FFTypeMethod.CGenFF_2b6)
dihedrals = mol.getSoftDihedrals()
for d in dihedrals:
print("\nScanning dihedral %s-%s-%s-%s" % (mol.name[d[0]], mol.name[d[1]], mol.name[d[2]], mol.name[d[3]]))
qmset = mol.scanSoftDihedral(d, directory="scan", step=20)
for i in range(len(qmset)):
print("%f %f" % (qmset[i].phi, qmset[i].energy))
def _analyse(self, mol, pdb, rtf, prm, traj, ftraj):
t = Molecule(pdb)
t.read(traj)
t.filter('not water')
t.write(ftraj)
m = FFMolecule(filename=mol, rtf=rtf, prm=prm)
m.read(ftraj)
torsions = m.getRotatableDihedrals()
# For each torsion
for i in range(len(torsions)):
# Create title
title = '{}-{}-{}-{}'.format(m.name[torsions[i][0]], m.name[torsions[i][1]], m.name[torsions[i][2]],
m.name[torsions[i][3]])
# Measure
(r, theta) = self._measure_torsion(torsions[i], m.coords)
self._plot_scatter(r, theta, title)
self._plot_hist(theta, title)
def main_scan():
if len(sys.argv) != 2:
print("Syntax: %s input.mol" % (sys.argv[0]))
sys.exit(0)
filename = sys.argv[1]
print("Scan {}".format(filename))
mol = FFMolecule(filename=filename, method=FFTypeMethod.CGenFF_2b6)
dihedrals = mol.getSoftDihedrals()
for d in dihedrals:
print("\nScanning dihedral %s-%s-%s-%s" %
(mol.name[d[0]], mol.name[d[1]], mol.name[d[2]], mol.name[d[3]]))
qmset = mol.scanSoftDihedral(d, directory="scan", step=20)
for i in range(len(qmset)):
print("%f %f" % (qmset[i].phi, qmset[i].energy))
def setUp(self):
from htmd.home import home
from htmd.parameterization.ffmolecule import FFMolecule, FFTypeMethod
molFile = os.path.join(home('test-param'), 'H2O2.mol2')
self.mol = FFMolecule(molFile, method=FFTypeMethod.GAFF2)
self.esp = ESP()
self.esp.molecule = self.mol
def test_getEquivalentDihedrals(self):
from htmd.home import home
from htmd.parameterization.ffmolecule import FFMolecule, FFTypeMethod
molFile = os.path.join(home('test-param'), 'glycol.mol2')
self.df.molecule = FFMolecule(molFile, method=FFTypeMethod.GAFF2)
self.assertListEqual(self.df._getEquivalentDihedrals([0, 1, 2, 3]),
[[0, 1, 2, 3]])
self.assertListEqual(self.df._getEquivalentDihedrals([4, 0, 1, 2]),
[[4, 0, 1, 2]])
self.assertListEqual(self.df._getEquivalentDihedrals([5, 1, 2, 7]),
[[5, 1, 2, 7]])
from tempfile import TemporaryDirectory
from htmd.home import home
from htmd.parameterization.ffmolecule import FFMolecule
from htmd.parameterization.fftype import FFTypeMethod
np.random.seed(20170801) # Make the tests deterministic
molFile = os.path.join(home('building-protein-ligand'), 'benzamidine.mol2')
methods = (FFTypeMethod.CGenFF_2b6, FFTypeMethod.GAFF, FFTypeMethod.GAFF2)
# TODO: remove then MATCH is fixed on Mac
methods = methods[1:] if os.environ.get(
'TRAVIS_OS_NAME') == 'osx' else methods
for method in methods:
mol = FFMolecule(molFile, method=method)
# Generate random charges
for name in mol._rtf.charge_by_name:
mol._rtf.charge_by_name[name] = 0.1 * np.random.randn()
# Generate a list of original and randomly distorted coordinates
coords = mol.coords[:, :, 0]
coordsList = [coords] + [
coords + 0.01 * np.random.randn(*coords.shape) for _ in range(9)
]
for coords in coordsList:
with TemporaryDirectory() as tmpDir:
import os
from tempfile import TemporaryDirectory
from htmd.home import home
from htmd.parameterization.ffmolecule import FFMolecule
from htmd.parameterization.fftype import FFTypeMethod
np.random.seed(20170801) # Make the tests deterministic
molFile = os.path.join(home('building-protein-ligand'), 'benzamidine.mol2')
methods = (FFTypeMethod.CGenFF_2b6, FFTypeMethod.GAFF, FFTypeMethod.GAFF2)
# TODO: remove then MATCH is fixed on Mac
methods = methods[1:] if os.environ.get('TRAVIS_OS_NAME') == 'osx' else methods
for method in methods:
mol = FFMolecule(molFile, method=method)
# Generate random charges
for name in mol._rtf.charge_by_name:
mol._rtf.charge_by_name[name] = 0.1*np.random.randn()
# Generate a list of original and randomly distorted coordinates
coords = mol.coords[:, :, 0]
coordsList = [coords] + [coords + 0.01*np.random.randn(*coords.shape) for _ in range(9)]
for coords in coordsList:
with TemporaryDirectory() as tmpDir:
if method == FFTypeMethod.CGenFF_2b6:
psfFile = os.path.join(tmpDir, 'mol.psf')
def main_parameterize(arguments=None):
args = getArgumentParser().parse_args(args=arguments)
if not os.path.exists(args.filename):
raise ValueError('File %s cannot be found' % args.filename)
method_map = {'GAFF': FFTypeMethod.GAFF, 'GAFF2': FFTypeMethod.GAFF2, 'CGENFF': FFTypeMethod.CGenFF_2b6}
methods = [method_map[method] for method in args.forcefield] # TODO: move into FFMolecule
# Get RTF and PRM file names
rtf, prm = None, None
if args.rtf_prm:
rtf, prm = args.rtf_prm
# Create a queue for QM
if args.queue == 'local':
queue = LocalCPUQueue()
elif args.queue == 'Slurm':
queue = SlurmQueue(_configapp=args.code.lower())
elif args.queue == 'LSF':
queue = LsfQueue(_configapp=args.code.lower())
elif args.queue == 'PBS':
queue = PBSQueue() # TODO: configure
elif args.queue == 'AceCloud':
queue = AceCloudQueue() # TODO: configure
else:
raise NotImplementedError
# Override default ncpus
if args.ncpus:
logger.info('Overriding ncpus to {}'.format(args.ncpus))
queue.ncpu = args.ncpus
# Create a QM object
if args.code == 'Psi4':
qm = Psi4()
elif args.code == 'Gaussian':
qm = Gaussian()
else:
raise NotImplementedError
# This is for debugging only!
if args.fake_qm:
qm = FakeQM()
logger.warning('Using FakeQM')
# Set up the QM object
qm.theory = args.theory
qm.basis = args.basis
qm.solvent = args.environment
qm.queue = queue
# List rotatable dihedral angles
if args.list:
mol = FFMolecule(args.filename, method=methods[0], netcharge=args.charge, rtf=rtf, prm=prm, qm=qm,
outdir=args.outdir)
print('\n === Parameterizable dihedral angles of %s ===\n' % args.filename)
with open('torsions.txt', 'w') as fh:
for dihedral in mol.getRotatableDihedrals():
dihedral_name = '%s-%s-%s-%s' % tuple(mol.name[dihedral])
print(' '+dihedral_name)
fh.write(dihedral_name+'\n')
print()
sys.exit(0)
# Print arguments
print('\n === Arguments ===\n')
for key, value in vars(args).items():
print('{:>12s}: {:s}'.format(key, str(value)))
print('\n === Parameterizing %s ===\n' % args.filename)
for method in methods:
print(" === Fitting for %s ===\n" % method.name)
# Create the molecule
mol = FFMolecule(args.filename, method=method, netcharge=args.charge, rtf=rtf, prm=prm, qm=qm,
outdir=args.outdir)
mol.printReport()
# Copy the molecule to preserve initial coordinates
mol_orig = mol.copy()
# Update B3LYP to B3LYP-D3
# TODO: this is silent and not documented stuff
if qm.theory == 'B3LYP':
qm.correction = 'D3'
# Update basis sets
# TODO: this is silent and not documented stuff
if mol.netcharge < 0 and qm.solvent == 'vacuum':
if qm.basis == '6-31G*':
qm.basis = '6-31+G*'
if qm.basis == 'cc-pVDZ':
qm.basis = 'aug-cc-pVDZ'
logger.info('Changing basis sets to %s' % qm.basis)
# Minimize molecule
if args.minimize:
print('\n == Minimizing ==\n')
mol.minimize()
# Fit ESP charges
if args.fit_charges:
print('\n == Fitting ESP charges ==\n')
# Set random number generator seed
if args.seed:
np.random.seed(args.seed)
# Select the atoms with fixed charges
fixed_atom_indices = []
for fixed_atom_name in args.fix_charge:
if fixed_atom_name not in mol.name:
raise ValueError('Atom %s is not found. Check --fix-charge arguments' % fixed_atom_name)
for aton_index in range(mol.numAtoms):
if mol.name[aton_index] == fixed_atom_name:
fixed_atom_indices.append(aton_index)
logger.info('Charge of atom %s is fixed to %f' % (fixed_atom_name, mol.charge[aton_index]))
# Fit ESP charges
score, qm_dipole = mol.fitCharges(fixed=fixed_atom_indices)
# Print results
mm_dipole = mol.getDipole()
score = np.sum((qm_dipole[:3] - mm_dipole[:3])**2)
print('Charge fitting score: %f\n' % score)
print('QM dipole: %f %f %f; %f' % tuple(qm_dipole))
print('MM dipole: %f %f %f; %f' % tuple(mm_dipole))
print('Dipole Chi^2 score: %f\n' % score)
# Fit dihedral angle parameters
if args.fit_dihedral:
print('\n == Fitting dihedral angle parameters ==\n')
# Set random number generator seed
if args.seed:
np.random.seed(args.seed)
# Get all rotatable dihedrals
all_dihedrals = mol.getRotatableDihedrals()
# Choose which dihedrals to fit
dihedrals = []
all_dihedral_names = ['-'.join(mol.name[dihedral]) for dihedral in all_dihedrals]
for dihedral_name in args.dihedral:
if dihedral_name not in all_dihedral_names:
raise ValueError('%s is not recognized as a rotatable dihedral angle' % dihedral_name)
dihedrals.append(all_dihedrals[all_dihedral_names.index(dihedral_name)])
dihedrals = dihedrals if len(dihedrals) > 0 else all_dihedrals # Set default to all dihedral angles
# Fit the parameters
mol.fitDihedrals(dihedrals, args.optimize_dihedral)
# Output the FF parameters
print('\n == Writing results ==\n')
mol.writeParameters(mol_orig)
# Write energy file
energyFile = os.path.join(mol.outdir, 'parameters', method.name, mol.output_directory_name(), 'energies.txt')
printEnergies(mol, energyFile)
logger.info('Write energy file: %s' % energyFile)
def main_parameterize():
ncpus = os.cpu_count()
try:
ncpus = int(os.getenv("NCPUS"))
except:
pass
parser = argparse.ArgumentParser(description="Acellera Small Molecule Parameterisation Version 2.0")
parser.add_argument("-m", "--mol2", help="Molecule to parameterise, in mol2 format", required=True, type=str,
default=None, metavar="<input.mol2>", action="store", dest="mol")
parser.add_argument("-l", "--list", "--list-torsions", help="List parameterisable torsions", action="store_true", default=False,
dest="list")
parser.add_argument("-c", "--charge", help="Net charge on molecule (default: sum of the partial charges on the "
".mol2 file)", type=int, default=None, action="store", dest="charge")
parser.add_argument("--rtf", help="Inital RTF parameters (req --prm)", type=str, default=None, dest="rtf")
parser.add_argument("--prm", help="Inital PRM parameters (req --rtf)", type=str, default=None, dest="prm")
parser.add_argument("-o", "--outdir", help="Output directory (default: %(default)s)", type=str, default="./",
dest="outdir")
parser.add_argument("-t", "--torsion", metavar="A1-A2-A3-A4", help="Torsion to parameterise (default: %(default)s)",
default="all", dest="torsion")
parser.add_argument("-n", "--ncpus", help="Number of CPUs to use (default: %(default)s)", default=ncpus,
dest="ncpus")
parser.add_argument("-f", "--forcefield", help="Inital FF guess to use (default: %(default)s)",
choices=["GAFF", "GAFF2", "CGENFF", "all"], default="all")
parser.add_argument("-b", "--basis", help="QM Basis Set (default: %(default)s)", choices=["6-31g-star", "cc-pVDZ"],
default="cc-pVDZ", dest="basis")
parser.add_argument("--theory", help="QM Theory (default: %(default)s)", choices=["RHF", "B3LYP"],
default="B3LYP", dest="theory")
parser.add_argument("--vacuum", help="Perform QM calculations in vacuum (default: %(default)s)",
action="store_true", dest="vacuum",
default=False)
parser.add_argument("--no-min", help="Do not perform QM minimisation (default: %(default)s)", action="store_true",
dest="nomin", default=False)
parser.add_argument("--no-esp", help="Do not perform QM charge fitting (default: %(default)s)", action="store_true",
dest="noesp", default=False)
parser.add_argument("--no-torsions", help="Do not perform torsion fitting (default: %(default)s)",
action="store_true", dest="notorsion", default=False)
parser.add_argument("-e", "--exec", help="Mode of execution for the QM calculations (default: %(default)s)",
choices=["inline", "LSF", "PBS", "Slurm", "AceCloud"], default="inline", dest="exec")
parser.add_argument("--qmcode", help="QM code (default: %(default)s)", choices=["Gaussian", "PSI4", "TeraChem"],
default="PSI4", dest="qmcode")
parser.add_argument("--freeze-charge", metavar="A1",
help="Freeze the charge of the named atom (default: %(default)s)", action="append",
default=None, dest="freezeq")
args = parser.parse_args()
# Communicate the # of CPUs to use to the QM engine via environment variable
os.environ['NCPUS'] = str(args.ncpus)
filename = args.mol
if not os.path.exists(filename):
print("File {} not found. Please check that the file exists and that the path is correct.".format(filename))
sys.exit(0)
if args.qmcode == "Gaussian":
code = Code.Gaussian
elif args.qmcode == "PSI4":
code = Code.PSI4
elif args.qmcode == "TeraChem":
code = Code.TeraChem
else:
print("Unknown QM code: {}".format(args.qmcode))
sys.exit(1)
if args.exec == "inline":
execution = Execution.Inline
elif args.exec == "LSF":
execution = Execution.LSF
elif args.exec == "PBS":
execution = Execution.PBS
elif args.exec == "Slurm":
execution = Execution.Slurm
elif args.exec == "AceCloud":
execution = Execution.AceCloud
else:
print("Unknown execution mode: {}".format(args.exec))
sys.exit(1)
if args.forcefield == "CGENFF":
methods = [FFTypeMethod.CGenFF_2b6]
elif args.forcefield == "GAFF":
methods = [FFTypeMethod.GAFF]
elif args.forcefield == "GAFF2":
methods = [FFTypeMethod.GAFF2]
elif args.forcefield == "all":
methods = [FFTypeMethod.CGenFF_2b6, FFTypeMethod.GAFF2]
else:
print("Unknown initial guess force-field: {}".format(args.forcefield))
sys.exit(1)
if args.basis == "6-31g-star":
basis = BasisSet._6_31G_star
elif args.basis == "cc-pVDZ":
basis = BasisSet._cc_pVDZ
else:
print("Unknown basis {}".format(args.basis))
sys.exit(1)
if args.theory == "RHF":
theory = Theory.RHF
elif args.theory == "B3LYP":
theory = Theory.B3LYP
else:
print("Unknown theory %s".format(args.theory))
sys.exit(1)
if args.vacuum:
solvent = False
else:
solvent = True
# Just list torsions?
if args.list:
print(" === Listing soft torsions of {} ===\n".format(filename))
mol = FFMolecule(filename=filename, method=methods[0], netcharge=args.charge, rtf=args.rtf, prm=args.prm,
basis=basis, theory=theory, solvent=solvent, execution=execution, qmcode=code,
outdir=args.outdir)
dihedrals = mol.getSoftTorsions()
print("Detected soft torsions:")
fh=open("torsions.txt", "w")
for d in dihedrals:
print("\t{}-{}-{}-{}".format(mol.name[d[0]], mol.name[d[1]], mol.name[d[2]], mol.name[d[3]]))
print("{}-{}-{}-{}".format(mol.name[d[0]], mol.name[d[1]], mol.name[d[2]], mol.name[d[3]]), file=fh)
fh.close()
sys.exit(0)
# Small report
print(" === List of arguments used ===\n")
for i in vars(args):
print('{:>10s}: {:<10s}'.format(i, str(vars(args)[i])))
print("\n === Parameterizing {} ===\n".format(filename))
for method in methods:
sys.stdout.flush()
print(" === Fitting for FF %s ===\n" % (method.name))
mol = FFMolecule(filename=filename, method=method, netcharge=args.charge, rtf=args.rtf, prm=args.prm,
basis=basis, theory=theory, solvent=solvent, execution=execution, qmcode=code,
outdir=args.outdir)
dihedrals = mol.getSoftTorsions()
if not args.nomin:
print("\n == Minimizing ==\n")
mol.minimize()
sys.stdout.flush()
if not args.noesp:
print("\n == Charge fitting ==\n")
# Select the atoms that are to have frozen charges in the fit
fixq = []
if args.freezeq:
for i in args.freezeq:
found = False
for d in range(len(mol.name)):
if mol.name[d] == i:
ni = d
print("Fixing charge for atom %s to %f" % (i, mol.charge[ni]))
fixq.append(ni)
found = True
if not found:
raise ValueError(" No atom named %s (--freeze-charge)" % i)
(score, qm_dipole, mm_dipole) = mol.fitCharges(fixed=fixq)
rating = "GOOD"
if score > 1:
rating = "CHECK"
if score > 10:
rating = "BAD"
print("Charge Chi^2 score : %f : %s" % (score, rating))
print("QM Dipole : %f %f %f ; %f" % (qm_dipole[0], qm_dipole[1], qm_dipole[2], qm_dipole[3]))
print("MM Dipole : %f %f %f ; %f" % (mm_dipole[0], mm_dipole[1], mm_dipole[2], mm_dipole[3]))
d = 0.
for i in range(3):
x = qm_dipole[i] - mm_dipole[i]
d = d + x * x
rating = "GOOD"
if score > 1:
rating = "CHECK"
print("Dipole Chi^2 score : %f : %s" % (d, rating))
print("")
sys.stdout.flush()
# Iterative dihedral fitting
if not args.notorsion:
print("\n == Torsion fitting ==\n")
scores = np.zeros(len(dihedrals))
converged = False
iteration = 1
while not converged:
rets = []
print("\nIteration %d" % iteration)
last_scores = scores
scores = np.zeros(len(dihedrals))
idx = 0
for d in dihedrals:
name = "%s-%s-%s-%s" % (mol.name[d[0]], mol.name[d[1]], mol.name[d[2]], mol.name[d[3]])
if args.torsion == 'all' or name in args.torsion.split(','):
print("\n == Fitting torsion {} ==\n".format(name))
try:
ret = mol.fitSoftTorsion(d)
rets.append(ret)
rating = "GOOD"
if ret.chisq > 10:
rating = "CHECK"
if ret.chisq > 100:
rating = "BAD"
print("Torsion %s Chi^2 score : %f : %s" % (name, ret.chisq, rating))
sys.stdout.flush()
scores[idx] = ret.chisq
fn = mol.plotTorsionFit(ret, show=False)
except:
print("Error in fitting")
# raise
scores[idx] = 0.
pass
# print(fn)
idx += 1
# print(scores)
if iteration > 1:
converged = True
for j in range(len(scores)):
# Check convergence
relerr = (scores[j] - last_scores[j]) / last_scores[j]
convstr = "- converged"
if math.fabs(relerr) > 1.e-2:
convstr = ""
converged = False
print(" Dihedral %d relative error : %f %s" % (j, relerr, convstr))
iteration += 1
print(" Fitting converged at iteration %d" % (iteration - 1))
fit = mol.plotConformerEnergies(rets, show=False)
print("\n Fit of conformer energies: RMS %f Variance %f" % (fit[0], fit[1]))
printEnergies(mol)
# Output the ff parameters
paramdir = os.path.join(args.outdir, "parameters", method.name, mol.output_directory_name())
print("\n == Output to {} ==\n".format(paramdir))
try:
os.makedirs(paramdir, exist_ok=True)
except:
raise OSError('Directory {} could not be created. Check if you have permissions.'.format(paramdir))
if method.name == "CGenFF_2b6":
try:
mol._rtf.write(os.path.join(paramdir, "mol.rtf"))
mol._prm.write(os.path.join(paramdir, "mol.prm"))
for ext in ['psf', 'xyz', 'coor', 'mol2', 'pdb']:
mol.write(os.path.join(paramdir, "mol." + ext))
f = open(os.path.join(paramdir, "input.namd"), "w")
tmp = '''parameters mol.prm
paraTypeCharmm on
coordinates mol.pdb
bincoordinates mol.coor
temperature 0
timestep 0
1-4scaling 1.0
exclude scaled1-4
outputname .out
outputenergies 1
structure mol.psf
cutoff 20.
switching off
stepsPerCycle 1
rigidbonds none
cellBasisVector1 50. 0. 0.
cellBasisVector2 0. 50. 0.
cellBasisVector3 0. 0. 50.
run 0'''
print(tmp, file=f)
f.close()
except ValueError as e:
print("Not writing CHARMM PRM: {}".format(str(e)))
elif method.name == "GAFF" or method.name == "GAFF2":
try:
# types need to be remapped because Amber FRCMOD format limits the type to characters
# writeFrcmod does this on the fly and returns a mapping that needs to be applied to the mol
typemap = mol._prm.writeFrcmod(mol._rtf, os.path.join(paramdir, "mol.frcmod"))
for ext in ['coor', 'mol2', 'pdb']:
mol.write(os.path.join(paramdir, "mol." + ext), typemap=typemap)
f = open(os.path.join(paramdir, "tleap.in"), "w")
tmp = '''loadAmberParams mol.frcmod
A = loadMol2 mol.mol2
saveAmberParm A structure.prmtop mol.crd
quit'''
print(tmp, file=f)
f.close()
f = open(os.path.join(paramdir, "input.namd"), "w")
tmp = '''parmfile structure.prmtop
amber on
coordinates mol.pdb
bincoordinates mol.coor
temperature 0
timestep 0
1-4scaling 0.83333333
exclude scaled1-4
outputname .out
outputenergies 1
cutoff 20.
switching off
stepsPerCycle 1
rigidbonds none
cellBasisVector1 50. 0. 0.
cellBasisVector2 0. 50. 0.
cellBasisVector3 0. 0. 50.
run 0'''
print(tmp, file=f)
f.close()
except ValueError as e:
print("Not writing Amber FRCMOD: {}".format(str(e)))
sys.exit(0)
if __name__ == '__main__':
import sys
import re
from htmd.home import home
from htmd.parameterization.ffmolecule import FFMolecule
# BUG: MATCH does not work on Mac!
if 'TRAVIS_OS_NAME' in os.environ:
if os.environ['TRAVIS_OS_NAME'] == 'osx':
sys.exit(0)
molFile = os.path.join(home('building-protein-ligand'), 'benzamidine.mol2')
refDir = home(dataDir='test-fftype/benzamidine')
mol = FFMolecule(molFile)
with TemporaryDirectory() as tmpDir:
ff = FFType(mol, method=FFTypeMethod.CGenFF_2b6)
ff._rtf.write(os.path.join(tmpDir, 'cgenff.rtf'))
ff._prm.write(os.path.join(tmpDir, 'cgenff.prm'))
ff = FFType(mol, method=FFTypeMethod.GAFF)
ff._prm.writeFrcmod(ff._rtf, os.path.join(tmpDir, 'gaff.frcmod'))
ff = FFType(mol, method=FFTypeMethod.GAFF2)
ff._prm.writeFrcmod(ff._rtf, os.path.join(tmpDir, 'gaff2.frcmod'))
for testFile in os.listdir(refDir):
print(testFile)
def main_parameterize(arguments=None):
args = getArgumentParser().parse_args(args=arguments)
if not os.path.exists(args.filename):
raise ValueError('File %s cannot be found' % args.filename)
method_map = {
'GAFF': FFTypeMethod.GAFF,
'GAFF2': FFTypeMethod.GAFF2,
'CGENFF': FFTypeMethod.CGenFF_2b6
}
methods = [method_map[method]
for method in args.forcefield] # TODO: move into FFMolecule
# Get RTF and PRM file names
rtf, prm = None, None
if args.rtf_prm:
rtf, prm = args.rtf_prm
# Create a queue for QM
if args.queue == 'local':
queue = LocalCPUQueue()
elif args.queue == 'Slurm':
queue = SlurmQueue(_configapp=args.code.lower())
elif args.queue == 'LSF':
queue = LsfQueue(_configapp=args.code.lower())
elif args.queue == 'PBS':
queue = PBSQueue() # TODO: configure
elif args.queue == 'AceCloud':
queue = AceCloudQueue() # TODO: configure
queue.groupname = args.groupname
queue.hashnames = True
else:
raise NotImplementedError
# Override default ncpus
if args.ncpus:
logger.info('Overriding ncpus to {}'.format(args.ncpus))
queue.ncpu = args.ncpus
if args.memory:
logger.info('Overriding memory to {}'.format(args.memory))
queue.memory = args.memory
# Create a QM object
if args.code == 'Psi4':
qm = Psi4()
elif args.code == 'Gaussian':
qm = Gaussian()
else:
raise NotImplementedError
# This is for debugging only!
if args.fake_qm:
qm = FakeQM2()
logger.warning('Using FakeQM')
# Set up the QM object
qm.theory = args.theory
qm.basis = args.basis
qm.solvent = args.environment
qm.queue = queue
# List rotatable dihedral angles
if args.list:
mol = FFMolecule(args.filename,
method=methods[0],
netcharge=args.charge,
rtf=rtf,
prm=prm,
qm=qm,
outdir=args.outdir)
print('\n === Parameterizable dihedral angles of %s ===\n' %
args.filename)
with open('torsions.txt', 'w') as fh:
for dihedral in mol.getRotatableDihedrals():
dihedral_name = '%s-%s-%s-%s' % tuple(mol.name[dihedral])
print(' ' + dihedral_name)
fh.write(dihedral_name + '\n')
print()
sys.exit(0)
# Print arguments
print('\n === Arguments ===\n')
for key, value in vars(args).items():
print('{:>12s}: {:s}'.format(key, str(value)))
print('\n === Parameterizing %s ===\n' % args.filename)
for method in methods:
print(" === Fitting for %s ===\n" % method.name)
# Create the molecule
mol = FFMolecule(args.filename,
method=method,
netcharge=args.charge,
rtf=rtf,
prm=prm,
qm=qm,
outdir=args.outdir)
mol.printReport()
# Copy the molecule to preserve initial coordinates
mol_orig = mol.copy()
# Update B3LYP to B3LYP-D3
# TODO: this is silent and not documented stuff
if qm.theory == 'B3LYP':
qm.correction = 'D3'
# Update basis sets
# TODO: this is silent and not documented stuff
if mol.netcharge < 0 and qm.solvent == 'vacuum':
if qm.basis == '6-31G*':
qm.basis = '6-31+G*'
if qm.basis == 'cc-pVDZ':
qm.basis = 'aug-cc-pVDZ'
logger.info('Changing basis sets to %s' % qm.basis)
# Minimize molecule
if args.minimize:
print('\n == Minimizing ==\n')
mol.minimize()
# Fit ESP charges
if args.fit_charges:
print('\n == Fitting ESP charges ==\n')
# Set random number generator seed
if args.seed:
np.random.seed(args.seed)
# Select the atoms with fixed charges
fixed_atom_indices = []
for fixed_atom_name in args.fix_charge:
if fixed_atom_name not in mol.name:
raise ValueError(
'Atom %s is not found. Check --fix-charge arguments' %
fixed_atom_name)
for aton_index in range(mol.numAtoms):
if mol.name[aton_index] == fixed_atom_name:
fixed_atom_indices.append(aton_index)
logger.info('Charge of atom %s is fixed to %f' %
(fixed_atom_name, mol.charge[aton_index]))
# Fit ESP charges
_, qm_dipole = mol.fitCharges(fixed=fixed_atom_indices)
# Copy the new charges to the original molecule
mol_orig.charge[:] = mol.charge
# Print dipoles
logger.info('QM dipole: %f %f %f; %f' % tuple(qm_dipole))
mm_dipole = mol.getDipole()
if np.all(np.isfinite(mm_dipole)):
logger.info('MM dipole: %f %f %f; %f' % tuple(mm_dipole))
else:
logger.warning(
'MM dipole cannot be computed. Check if elements are detected correctly.'
)
# Fit dihedral angle parameters
if args.fit_dihedral:
print('\n == Fitting dihedral angle parameters ==\n')
# Set random number generator seed
if args.seed:
np.random.seed(args.seed)
# Get all rotatable dihedrals
all_dihedrals = mol.getRotatableDihedrals()
# Choose which dihedrals to fit
dihedrals = []
all_dihedral_names = [
'-'.join(mol.name[dihedral]) for dihedral in all_dihedrals
]
for dihedral_name in args.dihedral:
if dihedral_name not in all_dihedral_names:
raise ValueError(
'%s is not recognized as a rotatable dihedral angle' %
dihedral_name)
dihedrals.append(
all_dihedrals[all_dihedral_names.index(dihedral_name)])
dihedrals = dihedrals if len(
dihedrals
) > 0 else all_dihedrals # Set default to all dihedral angles
# Fit the parameters
mol.fitDihedrals(dihedrals, args.optimize_dihedral)
# Output the FF parameters
print('\n == Writing results ==\n')
mol.writeParameters(mol_orig)
# Write energy file
energyFile = os.path.join(mol.outdir, 'parameters', method.name,
mol.output_directory_name(), 'energies.txt')
printEnergies(mol, energyFile)
logger.info('Write energy file: %s' % energyFile)
def main_parameterize(arguments=None):
args = cli_parser().parse_args(args=arguments)
from htmd.parameterization.ffmolecule import FFMolecule, FFEvaluate
from htmd.parameterization.fftype import FFTypeMethod
from htmd.qm.qmcalculation import Theory, BasisSet, Execution, Code
import numpy as np
import math
def printEnergies(m):
print("\n == Diagnostic Energies == ")
ffe = FFEvaluate(m)
energies = ffe.evaluate(m.coords[:, :, 0])
print("")
print(" Bond : %f" % (energies['bond']))
print(" Angle : %f" % (energies['angle']))
print(" Dihedral : %f" % (energies['dihedral']))
print(" Improper : %f" % (energies['improper']))
print(" Electro : %f" % (energies['elec']))
print(" VdW : %f" % (energies['vdw']))
print("")
# Communicate the # of CPUs to use to the QM engine via environment variable
os.environ['NCPUS'] = str(args.ncpus)
filename = args.mol
if not os.path.exists(filename):
print(
"File {} not found. Please check that the file exists and that the path is correct."
.format(filename))
sys.exit(0)
if args.qmcode == "Gaussian":
code = Code.Gaussian
elif args.qmcode == "PSI4":
code = Code.PSI4
elif args.qmcode == "TeraChem":
code = Code.TeraChem
else:
print("Unknown QM code: {}".format(args.qmcode))
sys.exit(1)
if args.exec == "inline":
execution = Execution.Inline
elif args.exec == "LSF":
execution = Execution.LSF
elif args.exec == "Slurm":
execution = Execution.Slurm
else:
print("Unknown execution mode: {}".format(args.exec))
sys.exit(1)
if args.forcefield == "CGENFF":
methods = [FFTypeMethod.CGenFF_2b6]
elif args.forcefield == "GAFF":
methods = [FFTypeMethod.GAFF]
elif args.forcefield == "GAFF2":
methods = [FFTypeMethod.GAFF2]
elif args.forcefield == "all":
methods = [FFTypeMethod.CGenFF_2b6, FFTypeMethod.GAFF2]
else:
print("Unknown initial guess force-field: {}".format(args.forcefield))
sys.exit(1)
if args.basis == "6-31g-star":
basis = BasisSet._6_31G_star
elif args.basis == "cc-pVDZ":
basis = BasisSet._cc_pVDZ
else:
print("Unknown basis {}".format(args.basis))
sys.exit(1)
if args.theory == "RHF":
theory = Theory.RHF
elif args.theory == "B3LYP":
theory = Theory.B3LYP
else:
print("Unknown theory %s".format(args.theory))
sys.exit(1)
if args.vacuum:
solvent = False
else:
solvent = True
# Just list torsions?
if args.list:
print(" === Listing soft torsions of {} ===\n".format(filename))
mol = FFMolecule(filename=filename,
method=methods[0],
netcharge=args.charge,
rtf=args.rtf,
prm=args.prm,
basis=basis,
theory=theory,
solvent=solvent,
execution=execution,
qmcode=code,
outdir=args.outdir)
dihedrals = mol.getSoftTorsions()
print("Detected soft torsions:")
fh = open("torsions.txt", "w")
for d in dihedrals:
print("\t{}-{}-{}-{}".format(mol.name[d[0]], mol.name[d[1]],
mol.name[d[2]], mol.name[d[3]]))
print("{}-{}-{}-{}".format(mol.name[d[0]], mol.name[d[1]],
mol.name[d[2]], mol.name[d[3]]),
file=fh)
fh.close()
sys.exit(0)
# Small report
print(" === List of arguments used ===\n")
for i in vars(args):
print('{:>10s}: {:<10s}'.format(i, str(vars(args)[i])))
print("\n === Parameterizing {} ===\n".format(filename))
for method in methods:
sys.stdout.flush()
print(" === Fitting for FF %s ===\n" % method.name)
mol = FFMolecule(filename=filename,
method=method,
netcharge=args.charge,
rtf=args.rtf,
prm=args.prm,
basis=basis,
theory=theory,
solvent=solvent,
execution=execution,
qmcode=code,
outdir=args.outdir)
dihedrals = mol.getSoftTorsions()
mol_orig = mol.copy()
if not args.nomin:
print("\n == Minimizing ==\n")
mol.minimize()
sys.stdout.flush()
if not args.noesp:
print("\n == Charge fitting ==\n")
# Select the atoms that are to have frozen charges in the fit
fixq = []
if args.freezeq:
for i in args.freezeq:
found = False
for d in range(len(mol.name)):
if mol.name[d] == i:
ni = d
print("Fixing charge for atom %s to %f" %
(i, mol.charge[ni]))
fixq.append(ni)
found = True
if not found:
raise ValueError(
" No atom named %s (--freeze-charge)" % i)
(score, qm_dipole, mm_dipole) = mol.fitCharges(fixed=fixq)
rating = "GOOD"
if score > 1:
rating = "CHECK"
if score > 10:
rating = "BAD"
print("Charge Chi^2 score : %f : %s" % (score, rating))
print("QM Dipole : %f %f %f ; %f" %
(qm_dipole[0], qm_dipole[1], qm_dipole[2], qm_dipole[3]))
print("MM Dipole : %f %f %f ; %f" %
(mm_dipole[0], mm_dipole[1], mm_dipole[2], mm_dipole[3]))
d = 0.
for i in range(3):
x = qm_dipole[i] - mm_dipole[i]
d = d + x * x
rating = "GOOD"
if score > 1:
rating = "CHECK"
print("Dipole Chi^2 score : %f : %s" % (d, rating))
print("")
sys.stdout.flush()
# Iterative dihedral fitting
if not args.notorsion:
print("\n == Torsion fitting ==\n")
scores = np.ones(len(dihedrals))
converged = False
iteration = 1
ref_mm = dict()
while not converged:
rets = []
print("\nIteration %d" % iteration)
last_scores = scores
scores = np.zeros(len(dihedrals))
idx = 0
for d in dihedrals:
name = "%s-%s-%s-%s" % (mol.name[d[0]], mol.name[d[1]],
mol.name[d[2]], mol.name[d[3]])
if args.torsion == 'all' or name in args.torsion.split(
','):
print("\n == Fitting torsion {} ==\n".format(name))
try:
ret = mol.fitSoftTorsion(d, geomopt=args.geomopt)
rets.append(ret)
if iteration == 1:
ref_mm[name] = ret
rating = "GOOD"
if ret.chisq > 10:
rating = "CHECK"
if ret.chisq > 100:
rating = "BAD"
print("Torsion %s Chi^2 score : %f : %s" %
(name, ret.chisq, rating))
sys.stdout.flush()
scores[idx] = ret.chisq
# Always use the mm_orig from first iteration (unmodified)
ret.mm_original = ref_mm[name].mm_original
phi_original = ref_mm[name].phi
fn = mol.plotTorsionFit(ret,
phi_original,
show=False)
except Exception as e:
print("Error in fitting")
print(str(e))
raise
scores[idx] = 0.
pass
# print(fn)
idx += 1
# print(scores)
if iteration > 1:
converged = True
for j in range(len(scores)):
# Check convergence
try:
relerr = (scores[j] -
last_scores[j]) / last_scores[j]
except:
relerr = 0.
if math.isnan(relerr):
relerr = 0.
convstr = "- converged"
if math.fabs(relerr) > 1.e-2:
convstr = ""
converged = False
print(" Dihedral %d relative error : %f %s" %
(j, relerr, convstr))
iteration += 1
print(" Fitting converged at iteration %d" % (iteration - 1))
if len(rets):
fit = mol.plotConformerEnergies(rets, show=False)
print("\n Fit of conformer energies: RMS %f Variance %f" %
(fit[0], fit[1]))
printEnergies(mol)
# Output the ff parameters
paramdir = os.path.join(args.outdir, "parameters", method.name,
mol.output_directory_name())
print("\n == Output to {} ==\n".format(paramdir))
try:
os.makedirs(paramdir, exist_ok=True)
except:
raise OSError(
'Directory {} could not be created. Check if you have permissions.'
.format(paramdir))
if method.name == "CGenFF_2b6":
try:
mol._rtf.write(os.path.join(paramdir, "mol.rtf"))
mol._prm.write(os.path.join(paramdir, "mol.prm"))
for ext in ['psf', 'xyz', 'coor', 'mol2', 'pdb']:
mol.write(os.path.join(paramdir, "mol." + ext))
mol_orig.write(os.path.join(paramdir, "mol-orig.mol2"))
f = open(os.path.join(paramdir, "input.namd"), "w")
tmp = '''parameters mol.prm
paraTypeCharmm on
coordinates mol.pdb
bincoordinates mol.coor
temperature 0
timestep 0
1-4scaling 1.0
exclude scaled1-4
outputname .out
outputenergies 1
structure mol.psf
cutoff 20.
switching off
stepsPerCycle 1
rigidbonds none
cellBasisVector1 50. 0. 0.
cellBasisVector2 0. 50. 0.
cellBasisVector3 0. 0. 50.
run 0'''
print(tmp, file=f)
f.close()
except ValueError as e:
print("Not writing CHARMM PRM: {}".format(str(e)))
elif method.name == "GAFF" or method.name == "GAFF2":
try:
# types need to be remapped because Amber FRCMOD format limits the type to characters
# writeFrcmod does this on the fly and returns a mapping that needs to be applied to the mol
typemap = mol._prm.writeFrcmod(
mol._rtf, os.path.join(paramdir, "mol.frcmod"))
for ext in ['coor', 'mol2', 'pdb']:
mol.write(os.path.join(paramdir, "mol." + ext),
typemap=typemap)
mol_orig.write(os.path.join(paramdir, "mol-orig.mol2"),
typemap=typemap)
f = open(os.path.join(paramdir, "tleap.in"), "w")
tmp = '''loadAmberParams mol.frcmod
A = loadMol2 mol.mol2
saveAmberParm A structure.prmtop mol.crd
quit'''
print(tmp, file=f)
f.close()
f = open(os.path.join(paramdir, "input.namd"), "w")
tmp = '''parmfile structure.prmtop
amber on
coordinates mol.pdb
bincoordinates mol.coor
temperature 0
timestep 0
1-4scaling 0.83333333
exclude scaled1-4
outputname .out
outputenergies 1
cutoff 20.
switching off
stepsPerCycle 1
rigidbonds none
cellBasisVector1 50. 0. 0.
cellBasisVector2 0. 50. 0.
cellBasisVector3 0. 0. 50.
run 0'''
print(tmp, file=f)
f.close()
except ValueError as e:
print("Not writing Amber FRCMOD: {}".format(str(e)))
sys.exit(0)
