def init_histogram(self, **kwargs):
# needs to be done separately because we might need additional information
# after init (at least I cannot think of a better way...)
smin = kwargs.pop("smin", self.min_coordinates(padding=self.padding))
smax = kwargs.pop("smax", self.max_coordinates(padding=self.padding))
BINS = fixedwidth_bins(self.delta, smin, smax)
self.arange = zip(BINS["min"], BINS["max"])
self.bins = BINS["Nbins"]
# create empty grid with the right dimensions (and get the edges)
grid, edges = numpy.histogramdd(numpy.zeros((1, 3)), bins=self.bins, range=self.arange, normed=False)
grid *= 0.0
h = grid.copy()
self.grid = grid
self.edges = edges
self._h = h # temporary for accumulation
def density_from_Universe(
universe,
delta=1.0,
atomselection="name OH2",
metadata=None,
padding=2.0,
cutoff=0,
soluteselection=None,
verbosity=3,
use_kdtree=True,
**kwargs
):
"""Create a density grid from a MDAnalysis.Universe object.
density_from_dcd(universe, delta=1.0, atomselection='name OH2', ...) --> density
:Arguments:
universe
:class:`MDAnalysis.Universe` object with a trajectory
:Keywords:
atomselection
selection string (MDAnalysis syntax) for the species to be analyzed
["name OH2"]
delta
approximate bin size for the density grid in Angstroem (same in x,y,z)
(It is slightly adjusted when the box length is not an integer multiple
of delta.) [1.0]
metadata
dictionary of additional data to be saved with the object
padding
increase histogram dimensions by padding (on top of initial box size)
in Angstroem [2.0]
soluteselection
MDAnalysis selection for the solute, e.g. "protein" [``None``]
cutoff
With *cutoff*, select '<atomsel> NOT WITHIN <cutoff> OF <soluteselection>'
(Special routines that are faster than the standard AROUND selection) [0]
verbosity: int
level of chattiness; 0 is silent, 3 is verbose [3]
parameters
dict with some special parameters for :class:`~hop.sitemap.Density` (see doc)
kwargs
metadata, parameters are modified and passed on to :class:`~hop.sitemap.Density`
:Returns: :class:`hop.sitemap.Density`
"""
try:
universe.selectAtoms("all")
universe.trajectory.ts
except AttributeError:
errmsg = "density_from_Universe(): The universe must be a proper MDAnalysis.Universe instance."
logger.fatal(errmsg)
raise TypeError(errmsg)
u = universe
if cutoff > 0 and soluteselection is not None:
# special fast selection for '<atomsel> not within <cutoff> of <solutesel>'
notwithin_coordinates = notwithin_coordinates_factory(
u, atomselection, soluteselection, cutoff, use_kdtree=use_kdtree
)
def current_coordinates():
return notwithin_coordinates()
else:
group = u.selectAtoms(atomselection)
def current_coordinates():
return group.coordinates()
coord = current_coordinates()
logger.info(
"Selected %d atoms out of %d atoms (%s) from %d total.",
coord.shape[0],
len(u.selectAtoms(atomselection)),
atomselection,
len(u.atoms),
)
# mild warning; typically this is run on RMS-fitted trajectories and
# so the box information is rather meaningless
box, angles = u.trajectory.ts.dimensions[:3], u.trajectory.ts.dimensions[3:]
if tuple(angles) <> (90.0, 90.0, 90.0):
logger.warn("Non-orthorhombic unit-cell --- make sure that it has been remapped properly!")
# Make the box bigger to avoid as much as possible 'outlier'. This
# is important if the sites are defined at a high density: in this
# case the bulk regions don't have to be close to 1 * n0 but can
# be less. It's much more difficult to deal with outliers. The
# ideal solution would use images: implement 'looking across the
# periodic boundaries' but that gets complicate when the box
# rotates due to RMS fitting.
smin = numpy.min(coord, axis=0) - padding
smax = numpy.max(coord, axis=0) + padding
BINS = fixedwidth_bins(delta, smin, smax)
arange = zip(BINS["min"], BINS["max"])
bins = BINS["Nbins"]
# create empty grid with the right dimensions (and get the edges)
grid, edges = numpy.histogramdd(numpy.zeros((1, 3)), bins=bins, range=arange, normed=False)
grid *= 0.0
h = grid.copy()
pm = CustomProgressMeter(
u.trajectory.numframes,
interval=10,
format="Histograming %(other)d atoms in frame %(step)5d/%(numsteps)d [%(percentage)5.1f%%]\r",
)
for ts in u.trajectory:
coord = current_coordinates()
if len(coord) == 0:
continue
h[:], edges[:] = numpy.histogramdd(coord, bins=bins, range=arange, normed=False)
grid += h # accumulate average histogram
pm.echo(ts.frame, len(coord))
numframes = u.trajectory.numframes / u.trajectory.skip
grid /= float(numframes)
# pick from kwargs
metadata = kwargs.pop("metadata", {})
metadata["psf"] = u.filename # named psf for historical reasons
metadata["dcd"] = u.trajectory.filename # named dcd for historical reasons
metadata["atomselection"] = atomselection
metadata["numframes"] = numframes
metadata["totaltime"] = round(
u.trajectory.numframes
* u.trajectory.delta
* u.trajectory.skip_timestep
* hop.constants.get_conversion_factor("time", "AKMA", "ps"),
3,
)
metadata["dt"] = (
u.trajectory.delta * u.trajectory.skip_timestep * hop.constants.get_conversion_factor("time", "AKMA", "ps")
)
metadata["time_unit"] = "ps"
metadata["dcd_skip"] = u.trajectory.skip_timestep # frames
metadata["dcd_delta"] = u.trajectory.delta # in AKMA
if cutoff > 0 and soluteselection is not None:
metadata["soluteselection"] = soluteselection
metadata["cutoff"] = cutoff # in Angstrom
parameters = kwargs.pop("parameters", {})
parameters["isDensity"] = False # must override
# all other kwargs are discarded
# Density automatically converts histogram to density for isDensity=False
g = Density(grid=grid, edges=edges, unit=dict(length="Angstrom"), parameters=parameters, metadata=metadata)
logger.info("Histogram completed (initial density in Angstrom**-3)")
return g
def __init__(
self, psf, pdb, delta=1.0, atomselection="name OH2", metadata=None, padding=4.0, sigma=None, verbosity=3
):
"""Construct the density from psf and pdb and the atomselection.
DC = BfactorDensityCreator(psf, pdb, delta=<delta>, atomselection=<MDAnalysis selection>,
metadata=<dict>, padding=2, sigma=None)
density = DC.PDBDensity()
psf Charmm psf topology file
pdb PDB file
atomselection
selection string (MDAnalysis syntax) for the species to be analyzed
delta approximate bin size for the density grid (same in x,y,z)
(It is slightly adjusted when the box length is not an integer multiple
of delta.)
metadata
dictionary of additional data to be saved with the object
padding increase histogram dimensions by padding (on top of initial box size)
sigma width (in Angstrom) of the gaussians that are used to build up the
density; if None then uses B-factors from pdb
verbosity=int level of chattiness; 0 is silent, 3 is verbose
For assigning X-ray waters to MD densities one might have to use a sigma
of about 0.5 A to obtain a well-defined and resolved x-ray water density
that can be easily matched to a broader density distribution.
"""
from MDAnalysis import Universe
set_verbosity(verbosity) # set to 0 for no messages
u = Universe(psf, pdbfilename=pdb)
group = u.selectAtoms(atomselection)
coord = group.coordinates()
logger.info(
"BfactorDensityCreator: Selected %d atoms (%s) out of %d total.",
coord.shape[0],
atomselection,
len(u.atoms),
)
smin = numpy.min(coord, axis=0) - padding
smax = numpy.max(coord, axis=0) + padding
BINS = fixedwidth_bins(delta, smin, smax)
arange = zip(BINS["min"], BINS["max"])
bins = BINS["Nbins"]
# get edges by doing a fake run
grid, self.edges = numpy.histogramdd(numpy.zeros((1, 3)), bins=bins, range=arange, normed=False)
self.delta = numpy.diag(map(lambda e: (e[-1] - e[0]) / (len(e) - 1), self.edges))
self.midpoints = map(lambda e: 0.5 * (e[:-1] + e[1:]), self.edges)
self.origin = map(lambda m: m[0], self.midpoints)
numframes = 1
if sigma is None:
# histogram individually, and smear out at the same time
# with the appropriate B-factor
if numpy.any(group.bfactors == 0.0):
wmsg = "BfactorDensityCreator: Some B-factors are Zero."
warnings.warn(wmsg, category=hop.MissingDataWarning)
logger.warn(wmsg)
rmsf = Bfactor2RMSF(group.bfactors)
grid *= 0.0 # reset grid
self.g = self._smear_rmsf(coord, grid, self.edges, rmsf)
else:
# histogram 'delta functions'
grid, self.edges = numpy.histogramdd(coord, bins=bins, range=arange, normed=False)
logger.info("Histogrammed %6d atoms from pdb.", len(group.atoms))
# just a convolution of the density with a Gaussian
self.g = self._smear_sigma(grid, sigma)
try:
metadata["psf"] = psf
except TypeError:
metadata = dict(psf=psf)
metadata["pdb"] = pdb
metadata["atomselection"] = atomselection
metadata["numframes"] = numframes
metadata["sigma"] = sigma
self.metadata = metadata
# Density automatically converts histogram to density for isDensity=False
logger.info("BfactorDensityCreator: Histogram completed (initial density in Angstrom**-3)\n")
