def cluster_dataset(array_like, options=''):
'''cluster dataset
Parameters
----------
array_like : array_like
options : str, cpptraj options
Returns
-------
cluster index for each data point
Examples
--------
>>> import pytraj as pt
>>> import numpy as np
>>> array_like = np.random.randint(0, 10, 1000)
>>> data = pt.cluster.cluster_dataset(array_like, 'clusters 10 epsilon 3.0')
'''
import numpy as np
c_dslist = CpptrajDatasetList()
c_dslist.add('double', '__array_like')
c_dslist[0].resize(len(array_like))
c_dslist[0].values[:] = array_like
act = c_analysis.Analysis_Clustering()
command = 'data __array_like ' + options
act(command, dslist=c_dslist)
return np.array(c_dslist[-2])
def test_1(self):
# just need to install libcpptraj with openmp
# that's it
# export OMP_NUM_THREADS=1
# python ./test_openmp_0.py
# export OMP_NUM_THREADS=8
# python ./test_openmp_0.py
dslist = DatasetList()
dflist = DataFileList()
traj = pt.iterload(fn('Tc5b.x'), fn('Tc5b.top'))
dslist.add("coords", "test_traj")
dslist[0].top = traj.top
for i in range(45):
dslist[0].load(traj.filename)
act = Analysis_Rms2d()
with Timer() as t:
act("crdset test_traj rmsout ./output/_test_2drms_CRDtest.openmp.dat",
dslist=dslist,
dflist=dflist)
# make sure to reproduce cpptraj to avoif false-impression :D
import numpy as np
matout = dslist[-1].get_full_matrix()
tmp = np.loadtxt("./data/test_openmp.Tc5b.n_threads_1.dat",
skiprows=1,
usecols=range(1, dslist[0].size + 1))
cpp_save = tmp.flatten()
# use decimal = 3 to mathc cpptraj's format here
assert_almost_equal(cpp_save, matout, decimal=3)
def test_reference_2(self):
traj = pt.iterload(fn('tz2.nc'), fn('tz2.parm7'))
from pytraj.analysis.c_action.c_action import Action_Rmsd
from pytraj.datasets.c_datasetlist import DatasetList
act = Action_Rmsd()
ref = traj[2]
dslist = DatasetList()
dslist.add('ref_frame', 'myref')
dslist[-1].top = traj.top
dslist[-1].add_frame(ref)
act('myrmsd refindex 0 @CA', traj, top=traj.top, dslist=dslist)
aa_eq(pt.rmsd(traj, ref=traj[2], mask='@CA'), dslist[-1].values)
def test_reference_2(self):
traj = pt.iterload("./data/tz2.nc", "./data/tz2.parm7")
from pytraj.c_action.c_action import Action_Rmsd
from pytraj.datasets.c_datasetlist import DatasetList
act = Action_Rmsd()
ref = traj[2]
dslist = DatasetList()
dslist.add('ref_frame', 'myref')
dslist[-1].top = traj.top
dslist[-1].add_frame(ref)
act('myrmsd refindex 0 @CA', traj, top=traj.top, dslist=dslist)
aa_eq(pt.rmsd(traj, ref=traj[2], mask='@CA'),
dslist[-1].values)
def test_actionlist(self):
'''test_actionlist
'''
dslist = DatasetList()
actlist = ActionList()
traj = pt.iterload("./data/Tc5b.x", "./data/Tc5b.top")
mask_list = ['@CB @CA', '@CA @H']
for mask in mask_list:
actlist.add(CA.Action_Vector(), mask, traj.top, dslist=dslist)
actlist.compute(traj)
dslist2 = pt.calc_vector(traj, mask_list)
dslist4 = va.vector_mask(traj, mask_list)
dslist3_0 = pt.calc_vector(traj, mask_list[0])
dslist3_1 = pt.calc_vector(traj, mask_list[1])
aa_eq(dslist3_0, dslist2[0])
aa_eq(dslist3_1, dslist2[1])
aa_eq(dslist3_0, dslist4[0])
aa_eq(dslist3_1, dslist4[1])
aa_eq(dslist3_0, dslist[0].values)
aa_eq(dslist3_1, dslist[1].values)
def calc_dssp(traj=None,
mask="",
frame_indices=None,
dtype='ndarray',
simplified=False,
top=None):
"""return dssp profile for frame/traj
Parameters
----------
traj : Trajectory-like
mask: str
atom mask
frame_indices : {None, array-like}, default None, optional
specify frame numbers for calculation.
if None, do all frames
dtype : str, default 'ndarray'
return data type, for regular user, just use default one (ndarray).
use dtype='dataset' if wanting to get secondary structure in integer format
simplified : bool, default False
if True, use simplified codes, only has 'H', 'E' and 'C'
if False, use all DSSP codes
Returns
-------
out_0: ndarray, shape=(n_residues,)
residue names
out_1: ndarray, shape=(n_frames, n_residues)
DSSP for each residue
out_2 : pytraj.DatasetList
average value for each secondary structure type
Examples
--------
>>> import pytraj as pt
>>> traj = pt.load_pdb_rcsb('1l2y')
>>> residues, ss, _ = pt.dssp(traj, ":2-10")
>>> residues # doctest: +SKIP
array(['LEU:2', 'TYR:3', 'ILE:4', 'GLN:5', 'TRP:6', 'LEU:7', 'LYS:8',
'ASP:9', 'GLY:10'],
dtype='<U6')
>>> ss # doctest: +SKIP
array([['0', 'H', 'H', ..., 'H', 'T', '0'],
['0', 'H', 'H', ..., 'H', 'T', '0'],
['0', 'H', 'H', ..., 'H', 'T', '0'],
...,
['0', 'H', 'H', ..., 'H', 'T', '0'],
['0', 'H', 'H', ..., 'H', 'H', '0'],
['0', 'H', 'H', ..., 'H', 'T', '0']],
dtype='<U1')
>>> residues, ss, _ = pt.dssp(traj, mask=range(100))
>>> traj = pt.fetch_pdb('1l2y')
>>> residues, ss, _ = pt.dssp(traj, simplified=True)
>>> ss[0].tolist() # first frame
['C', 'H', 'H', 'H', 'H', 'H', 'H', 'H', 'C', 'C', 'H', 'H', 'H', 'H', 'C', 'C', 'C', 'C', 'C', 'C']
Notes
-----
========= ======= ========= =======================
Character Integer DSSP_Char Seconday structure type
========= ======= ========= =======================
0 0 '0' None
b 1 'E' Parallel Beta-sheet
B 2 'B' Anti-parallel Beta-sheet
G 3 'G' 3-10 helix
H 4 'H' Alpha helix
I 5 'I' Pi (3-14) helix
T 6 'T' Turn
S 7 'S' Bend
========= ======= ========= =======================
Simlified codes::
- 'H': include 'H', 'G', 'I' (helix)
- 'E': include 'E', 'B' (strand)
- 'C': include 'T', 'S' or '0' (coil)
Simlified codes will be mostly used for visulization in other packages.
"""
from pytraj.datasets.c_datasetlist import DatasetList as CpptrajDatasetList
from pytraj.c_action.c_action import Action_DSSP
command = mask
dslist = CpptrajDatasetList()
Action_DSSP()(command, traj, top=top, dslist=dslist)
# replace key to something nicer
for key, dset in dslist.iteritems():
if 'DSSP' in key:
key = key.replace("DSSP_00000[", "")
key = key.replace("]", "_avg")
dset.key = key.lower()
dtype = dtype.lower()
if dtype == 'ndarray':
# get all dataset from DatSetList if dtype == integer
arr0 = dslist.grep("integer", mode='dtype').values
keys = dslist.grep("integer", mode='dtype').keys()
avg_dict = DatasetList(dslist.grep('_avg'))
ss_array = np.asarray([
_to_string_secondary_structure(arr, simplified=simplified)
for arr in arr0
]).T
return np.asarray(keys), ss_array, avg_dict
else:
return get_data_from_dtype(dslist, dtype=dtype)
def calc_dssp(traj=None,
mask="",
frame_indices=None,
dtype='ndarray',
simplified=False,
top=None):
"""return dssp profile for frame/traj
Parameters
----------
traj : Trajectory-like
mask: str
atom mask
frame_indices : {None, array-like}, default None, optional
specify frame numbers for calculation.
if None, do all frames
dtype : str, default 'ndarray'
return data type, for regular user, just use default one (ndarray).
use dtype='dataset' if wanting to get secondary structure in integer format
simplified : bool, default False
if True, use simplified codes, only has 'H', 'E' and 'C'
if False, use all DSSP codes
Returns
-------
out_0: ndarray, shape=(n_residues,)
residue names
out_1: ndarray, shape=(n_frames, n_residues)
DSSP for each residue
out_2 : pytraj.DatasetList
average value for each secondary structure type
Examples
--------
>>> import pytraj as pt
>>> traj = pt.load_pdb_rcsb('1l2y')
>>> residues, ss, _ = pt.dssp(traj, ":2-10")
>>> residues # doctest: +SKIP
array(['LEU:2', 'TYR:3', 'ILE:4', 'GLN:5', 'TRP:6', 'LEU:7', 'LYS:8',
'ASP:9', 'GLY:10'],
dtype='<U6')
>>> ss # doctest: +SKIP
array([['0', 'H', 'H', ..., 'H', 'T', '0'],
['0', 'H', 'H', ..., 'H', 'T', '0'],
['0', 'H', 'H', ..., 'H', 'T', '0'],
...,
['0', 'H', 'H', ..., 'H', 'T', '0'],
['0', 'H', 'H', ..., 'H', 'H', '0'],
['0', 'H', 'H', ..., 'H', 'T', '0']],
dtype='<U1')
>>> residues, ss, _ = pt.dssp(traj, mask=range(100))
>>> traj = pt.fetch_pdb('1l2y')
>>> residues, ss, _ = pt.dssp(traj, simplified=True)
>>> ss[0].tolist() # first frame
['C', 'H', 'H', 'H', 'H', 'H', 'H', 'H', 'C', 'C', 'H', 'H', 'H', 'H', 'C', 'C', 'C', 'C', 'C', 'C']
Notes
-----
========= ======= ========= =======================
Character Integer DSSP_Char Seconday structure type
========= ======= ========= =======================
0 0 '0' None
b 1 'E' Parallel Beta-sheet
B 2 'B' Anti-parallel Beta-sheet
G 3 'G' 3-10 helix
H 4 'H' Alpha helix
I 5 'I' Pi (3-14) helix
T 6 'T' Turn
S 7 'S' Bend
========= ======= ========= =======================
Simlified codes::
- 'H': include 'H', 'G', 'I' (helix)
- 'E': include 'E', 'B' (strand)
- 'C': include 'T', 'S' or '0' (coil)
Simlified codes will be mostly used for visulization in other packages.
"""
from pytraj.datasets.c_datasetlist import DatasetList as CpptrajDatasetList
from pytraj.c_action.c_action import Action_DSSP
command = mask
dslist = CpptrajDatasetList()
Action_DSSP()(command, traj, top=top, dslist=dslist)
# replace key to something nicer
for key, dset in dslist.iteritems():
if 'DSSP' in key:
key = key.replace("DSSP_00000[", "")
key = key.replace("]", "_avg")
dset.key = key.lower()
dtype = dtype.lower()
if dtype == 'ndarray':
# get all dataset from DatSetList if dtype == integer
arr0 = dslist.grep("integer", mode='dtype').values
keys = dslist.grep("integer", mode='dtype').keys()
avg_dict = DatasetList(dslist.grep('_avg'))
ss_array = np.asarray([_to_string_secondary_structure(
arr,
simplified=simplified) for arr in arr0]).T
return np.asarray(keys), ss_array, avg_dict
else:
return get_data_from_dtype(dslist, dtype=dtype)
def energy_decomposition(traj=None,
prmtop=None,
igb=8,
mm_options=None,
qm_options=None,
mode=None,
dtype='dict',
frame_indices=None,
top=None):
"""energy decomposition by calling `libsander`
Parameters
----------
traj : Trajectory-like or iterables that produce Frame
if `traj` does not hold Topology information, `top` must be provided
prmtop : str or Structure from ParmEd, default=None, optional
To avoid any unexpected error, you should always provide original topology
filename. If prmtop is None, pytraj will load Topology from traj.top.filename.
- why do you need to load additional topology filename? Because cpptraj and sander
use different Topology object, can not convert from one to another.
igb : GB model, default=8 (GB-Neck2)
If specify `mm_options`, this `igb` input will be ignored
mm_options : InputOptions from `sander`, default=None, optional
if `mm_options` is None, use `gas_input` with given igb.
If `mm_options` is not None, use this
qm_options : InputOptions from `sander` for QMMM, optional
mode : str, default=None, optional
if mode='minimal', get only 'bond', 'angle', 'dihedral' and 'total' energies
top : pytraj.Topology or str, default=None, optional
only need to specify this ``top`` if ``traj`` does not hold Topology
dtype : str, {'dict', 'dataset', 'ndarray', 'dataframe'}, default='dict'
return data type
frame_indices : None or 1D array-like, default None
if not None, only perform calculation for given frames
Returns
-------
Dict of energies (to be used with DataFrame) or DatasetList
Examples
--------
Examples are adapted from $AMBERHOME/test/sanderapi
>>> import pytraj as pt
>>> # GB energy
>>> traj = pt.datafiles.load_ala3()
>>> traj.n_frames
1
>>> data = pt.energy_decomposition(traj, igb=8)
>>> data['gb']
array([-92.88577683])
>>> data['bond']
array([ 5.59350521])
>>> # PME
>>> import os
>>> from pytraj.testing import amberhome
>>> import sander
>>> topfile = os.path.join(amberhome, "test/4096wat/prmtop")
>>> rstfile = os.path.join(amberhome, "test/4096wat/eq1.x")
>>> traj = pt.iterload(rstfile, topfile)
>>> options = sander.pme_input()
>>> options.cut = 8.0
>>> edict = pt.energy_decomposition(traj=traj, mm_options=options)
>>> edict['vdw']
array([ 6028.95167558])
>>> # GB + QMMM
>>> topfile = os.path.join(amberhome, "test/qmmm2/lysine_PM3_qmgb2/prmtop")
>>> rstfile = os.path.join(amberhome, "test/qmmm2/lysine_PM3_qmgb2/lysine.crd")
>>> traj = pt.iterload(rstfile, topfile)
>>> options = sander.gas_input(8)
>>> options.cut = 99.0
>>> options.ifqnt = 1
>>> qm_options = sander.qm_input()
>>> qm_options.iqmatoms[:3] = [8, 9, 10]
>>> qm_options.qm_theory = "PM3"
>>> qm_options.qmcharge = 0
>>> qm_options.qmgb = 2
>>> qm_options.adjust_q = 0
>>> edict = pt.energy_decomposition(traj=traj, mm_options=options, qm_options=qm_options)
>>> edict['bond']
array([ 0.00160733])
>>> edict['scf']
array([-11.92177575])
Notes
-----
This method does not work with `pytraj.pmap` when you specify mm_options and
qm_options. Use `pytraj.pmap_mpi` with MPI instead.
Work with ``pytraj.pmap``::
pt.pmap(pt.energy_decomposition, traj, igb=8, dtype='dict')
Will NOT work with ``pytraj.pmap``::
import sander
inp = sander.gas_input(8)
pt.pmap(pt.energy_decomposition, traj, mm_options=inp, dtype='dict')
Why? Because Python need to pickle each object to send to different cores and Python
does not know how to pickle mm_options from sander.gas_input(8).
This works with ``pytraj.pmap_mpi`` because pytraj explicitly create ``mm_options``
in each core without pickling.
"""
from collections import defaultdict, OrderedDict
from pytraj.misc import get_atts
import numpy as np
try:
import sander
except ImportError:
raise ImportError("need both `pysander` installed. Check Ambertools15")
ddict = defaultdict(_default_func)
if mm_options is None:
inp = sander.gas_input(igb)
elif igb is not None:
inp = mm_options
if isinstance(inp, string_types):
# dangerous
local_dict = {'sander': sander}
exec(inp.lstrip(), local_dict)
inp = local_dict['mm_options']
if isinstance(qm_options, string_types):
# dangerous
local_dict = {'sander': sander}
exec(qm_options.lstrip(), local_dict)
qm_options = local_dict['qm_options']
if prmtop is None:
try:
# try to load from file by taking top.filename
prmtop_ = top.filename
except AttributeError:
raise ValueError("prmtop must be AmberParm object in ParmEd")
else:
# Structure, string
prmtop_ = prmtop
if not hasattr(prmtop_, 'coordinates') or prmtop_.coordinates is None:
try:
# if `traj` is Trajectory-like (not frame_iter), try to take 1st
# coords
coords = traj[0].xyz
except (TypeError, AttributeError):
# create fake list
coords = [0. for _ in range(top.n_atoms * 3)]
else:
# use default coords in `AmberParm`
coords = prmtop_.coordinates
if top.has_box():
box = top.box.tolist()
has_box = True
else:
box = None
has_box = False
with sander.setup(prmtop_, coords, box, inp, qm_options):
for frame in iterframe_master(traj):
if has_box:
sander.set_box(*frame.box.tolist())
sander.set_positions(frame.xyz)
ene, frc = sander.energy_forces()
# potentially slow
ene_atts = get_atts(ene)
for att in ene_atts:
ddict[att].append(getattr(ene, att))
new_dict = None
if mode == 'minimal':
new_dict = {}
for key in ['bond', 'angle', 'dihedral', 'tot']:
new_dict[key] = ddict[key]
else:
new_dict = ddict
for key in new_dict.keys():
new_dict[key] = np.asarray(new_dict[key])
if dtype == 'dict':
return OrderedDict(new_dict)
else:
from pytraj.datasets.c_datasetlist import DatasetList
dslist = DatasetList()
size = new_dict['tot'].__len__()
for key in new_dict.keys():
dslist.add('double')
dslist[-1].key = key
dslist[-1].resize(size)
dslist[-1].data[:] = new_dict[key]
return get_data_from_dtype(dslist, dtype)