Beispiel #1
0
min_crds = [coords[0][0], coords[0][1], coords[0][2]]
max_crds = [coords[0][0], coords[0][1], coords[0][2]]

for coord in coords:
    min_crds[0] = min(min_crds[0], coord[0])
    min_crds[1] = min(min_crds[1], coord[1])
    min_crds[2] = min(min_crds[2], coord[2])
    max_crds[0] = max(max_crds[0], coord[0])
    max_crds[1] = max(max_crds[1], coord[1])
    max_crds[2] = max(max_crds[2], coord[2])

psf_solvate_1 = CharmmPsfFile('../structure/butane_solvated.psf')
psf_solvate_1.box = (max_crds[0] - min_crds[0], max_crds[1] - min_crds[1],
                     max_crds[2] - min_crds[2], 90.0, 90.0, 90.0)
database['solvated']['structure_1'] = psf_solvate_1
psf_solvate_1.load_parameters(param_1, copy_parameters=False)
system_solvate_1 = psf_solvate_1.createSystem(
    nonbondedMethod=app.PME,
    constraints=app.HBonds,
    nonbondedCutoff=12.0 * u.angstroms,
    switchDistance=10.0 * u.angstroms)
database['solvated']['system_1'] = system_solvate_1

psf_solvate_2 = CharmmPsfFile('../structure/butane_solvated.psf')
psf_solvate_2.box = (max_crds[0] - min_crds[0], max_crds[1] - min_crds[1],
                     max_crds[2] - min_crds[2], 90.0, 90.0, 90.0)
database['solvated']['structure_2'] = psf_solvate_2
psf_solvate_2.load_parameters(param_2, copy_parameters=False)
system_solvate_2 = psf_solvate_2.createSystem(
    nonbondedMethod=app.PME,
    constraints=app.HBonds,
Beispiel #2
0
class ToyModel(object):
    """
    Toy model of 4 carbons to test torsionfit

    Attributes
    ----------
    synthetic_energy: simtk.units.Quantity((n_increments), unit=kilojoule/mole)
    true_value: parmed.topologyobjects.DihedralTypeList (the dihedral parameters used to calculate synthetic energy)
    initital_value: parmed.topologyobjects.DihedralTypeList (dihedral parameters used to initialize pymc model)
    torsion_scan: torsionfit.TorsionScanSet
    model: torsionfit.TorsionFitModel

    The default toy model has randomized true and initial dihedral parameters. This can be changed by passing a
    parmed.topologyobjects.DihedralTypeList to true_value and/or initial_value when initializing a model instance.

    """

    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 _spher2cart(self, r, theta, phi):
        """convert spherical to cartesian coordinates

        Paramters:
        r: bond length
        theta: bond angle
        phi: dihedral angle

        returns:
        cartesian coordinates
        """
        x = r * np.sin(theta) * np.cos(phi)
        y = r * np.sin(theta) * np.sin(phi)
        z = r * np.cos(theta)
        return [x, y, z]

    def _torsion_scan(self, n_increments):
        """
        Generate positions and energy for torsion scan
        Parameters:
            n_increments: int. how many points to scan
        """
        n_increments = n_increments
        n_atoms = 4
        phis = np.arange(-np.pi, +np.pi, np.pi/n_increments)
        positions = np.zeros((len(phis), n_atoms, 3))

        # Get bond length in nm
        for bond in self._struct.bonds:
            if bond.atom1.type == bond.atom2.type:
                length = units.Quantity(value=bond.type.req * 0.1, unit=units.nanometers) # convert to nm
        # Get angle in radians
        for angle in self._struct.angles:
            if angle.atom1.type == angle.atom2.type:
                theta = units.Quantity(value=angle.type.theteq * (np.pi/180.0), unit=units.radians)

        atom1_coords = self._spher2cart(length.value_in_unit(units.nanometer), theta.value_in_unit(units.radian), phis[0])
        for i, phi in enumerate(phis):
            atom3_coords = self._spher2cart(length.value_in_unit(units.nanometer), theta.value_in_unit(units.radian), phi)
            atom3_coords[-1] = abs(atom3_coords[-1]) + length._value
            positions[i] = [atom1_coords,
                [0.000, 0.000, 0.000],
                [0.000, 0.000, length._value],
                atom3_coords]
        self._positions = units.Quantity(value=positions, unit=units.nanometer) # put units back in

        # calculate energy
        self.synthetic_energy = units.Quantity(value=np.zeros((len(positions))), unit=units.kilojoules_per_mole)
        platform = mm.Platform.getPlatformByName('Reference')
        integrator = mm.VerletIntegrator(0.004*units.picoseconds)
        system = self._struct.createSystem()
        context = mm.Context(system, integrator, platform)
        for i, conf in enumerate(self._positions):
            context.setPositions(conf)
            state = context.getState(getEnergy=True)
            energy = state.getPotentialEnergy()
            self.synthetic_energy[i] = energy + units.Quantity(value=np.random.normal(0, 1.0), unit=units.kilojoules_per_mole)

    def _randomize_dih_param(self, return_dih=False):
        """
        generates random dihedral parameters

        """
        dih_tlist = DihedralTypeList()
        multiplicities = [1, 2, 3, 4, 6]
        terms = np.random.randint(1, 5+1)
        np.random.shuffle(multiplicities)
        for i in range(terms):
            k = np.random.uniform(0.0, 20.0)
            n = multiplicities[i]
            phase = np.random.randint(0, 1+1)
            if phase == 1:
                phase = 180
            dih_tlist.append(DihedralType(k, n, phase, 1.00, 1.00))
        self._param.dihedral_types[self._dih_type] = dih_tlist
        if return_dih:
            _dih_tlist = copy.deepcopy(dih_tlist)
            return _dih_tlist

    def save_torsion_values(self, filename=None):
        """

        Parameters
        ----------
        filename : str
            file to save torsion values. Default is None. If None, return np.array

        Returns
        -------
        dih_param: np.array
            array containing torison parameters
        """
        dih_param = np.ones(shape=(2, 6, 3)) * np.nan
        for i, dih in enumerate(self.true_value):
            dih_param[0, i, 0] = dih.per
            dih_param[0, i, 1] = dih.phi_k
            dih_param[0, i, 2] = dih.phase

        for j, dih in enumerate(self.initial_value):
            dih_param[1, j, 0] = dih.per
            dih_param[1, j, 1] = dih.phi_k
            dih_param[1, j, 2] = dih.phase

        if filename is not None:
            np.save(filename, dih_param)
        else:
            return dih_param

    @staticmethod
    def from_dih_params(filename=None, dih_params=None, rj=False, continuous=False, n_increments=13, sample_phase=False):
        """

        Parameters
        ----------
        filename : str
            name of file with serialized dihedral parameters
        rj : bool
            Flag if using reversible jump. Default it True
        continuous : bool
            Flag if sampling continuous phase. Default is False
        n_increments : int
            incermentation of torsion drive
        sample_phase : bool
            Flag if sampling phase. Default is False (K is allowed to go negative when sample_phase is False)

        Returns
        -------
        ToyModel with true and initial value from saved file.

        """
        if filename is None and dih_params is None:
            msg = 'You must provide either an npy file or a numpy array with true and initial values for the toy model'
            raise Exception(msg)
        if filename is not None:
            dih_params = np.load(filename)
        dih_tlist_true = DihedralTypeList()
        dih_tlist_init = DihedralTypeList()

        true = dih_params[0]
        init = dih_params[1]

        for dih in true:
            if not np.isnan(dih[0]):
                dih_tlist_true.append(DihedralType(per=dih[0], phi_k=dih[1], phase=dih[2]))

        for dih in init:
            if not np.isnan(dih[0]):
                dih_tlist_init.append(DihedralType(per=dih[0], phi_k=dih[1], phase=dih[2]))

        return ToyModel(true_value=dih_tlist_true, initial_value=dih_tlist_init, rj=rj, continuous=continuous,
                        n_increments=n_increments, sample_phase=sample_phase)