def shukla_coords(trajectories,KER,Aloop,SRC2):
difference = []
rmsd = []
for traj in trajectories:
# append difference
k295e310 = md.compute_contacts(traj, [KER[0]])
e310r409 = md.compute_contacts(traj, [KER[1]])
difference.append(10*(e310r409[0] - k295e310[0])) # 10x because mdtraj is naturally in nm
# append rmsd
Activation_Loop_SRC2 = SRC2.top.select("backbone and (resid %s to %s)" %(140,160))
Activation_Loop_kinase = traj.top.select("backbone and (resid %s to %s)" %(Aloop[0],Aloop[1]))
SRC2_cut = SRC2.atom_slice(Activation_Loop_SRC2)
traj_cut = traj.atom_slice(Activation_Loop_kinase)
rmsd.append(10*(md.rmsd(traj_cut,SRC2_cut,frame=0))) # 10x because mdtraj is naturaly in nm
# flatten list of arrays
flattened_difference = np.asarray([val for sublist in difference for val in sublist])
flattened_rmsd = np.asarray([val for sublist in rmsd for val in sublist])
return [flattened_rmsd, flattened_difference]
def test_trek():
# setup
with open("processed/p9761/24/7/info.json") as f:
info = json.load(f)
info = trajprocess.postprocess.stp(info, 'trek')
# check stp cleanup
assert not os.path.exists('{workdir}/stp/0/'.format(**info['path']))
# check stp results
traj = mdtraj.load(info['stp']['gens'][0], top=info['stp']['outtop'])
assert traj.n_atoms == 30962
assert len(traj) == 7
# do ctr
info = trajprocess.postprocess.ctr(info, "trek")
# check ctr info
assert not os.path.exists("{workdir}/cpptraj.tmp".format(**info['path']))
assert not os.path.exists(
"{workdir}/ctr/cpptraj.tmp".format(**info['path']))
traj2 = mdtraj.load(info['ctr']['gens'][0], top=info['stp']['outtop'])
# check ctr results
# Trek has 518 protein residues
pairs = np.random.randint(0, 518, (20, 2))
cont1, _ = mdtraj.compute_contacts(traj, pairs)
cont2, _ = mdtraj.compute_contacts(traj2, pairs)
np.testing.assert_array_almost_equal(cont1, cont2, decimal=4)
def test_contact_0():
pdb = md.load(get_fn('bpti.pdb'))
contacts = np.loadtxt(get_fn('contacts.dat')).astype(int)
ca, ca_pairs = md.compute_contacts(pdb, contacts, scheme='ca')
closest, closest_pairs = md.compute_contacts(pdb, contacts, scheme='closest')
closest_heavy, closest_heavy_pairs = md.compute_contacts(pdb, contacts, scheme='closest-heavy')
sidechain, sidechain_pairs = md.compute_contacts(pdb, contacts, scheme='sidechain')
sidechain_heavy, sidechain_heavy_pairs = md.compute_contacts(pdb, contacts, scheme='sidechain-heavy')
ref_ca = np.loadtxt(get_fn('cc_ca.dat'))
ref_closest = np.loadtxt(get_fn('cc_closest.dat'))
ref_closest_heavy = np.loadtxt(get_fn('cc_closest-heavy.dat'))
ref_sidechain = np.loadtxt(get_fn('cc_sidechain.dat'))
ref_sidechain_heavy = np.loadtxt(get_fn('cc_sidechain-heavy.dat'))
eq(ref_ca, ca.flatten())
eq(ref_closest, closest.flatten())
eq(ref_closest_heavy, closest_heavy.flatten())
eq(ref_sidechain, sidechain.flatten())
eq(ref_sidechain_heavy, sidechain_heavy.flatten())
eq(contacts, ca_pairs)
eq(contacts, closest_pairs)
eq(contacts, closest_heavy_pairs)
eq(contacts, sidechain_pairs)
eq(contacts, sidechain_heavy_pairs)
def test_contact_0(get_fn):
pdb = md.load(get_fn('bpti.pdb'))
contacts = np.loadtxt(get_fn('contacts.dat')).astype(int)
ca, ca_pairs = md.compute_contacts(pdb, contacts, scheme='ca')
closest, closest_pairs = md.compute_contacts(pdb,
contacts,
scheme='closest')
closest_heavy, closest_heavy_pairs = md.compute_contacts(
pdb, contacts, scheme='closest-heavy')
sidechain, sidechain_pairs = md.compute_contacts(pdb,
contacts,
scheme='sidechain')
sidechain_heavy, sidechain_heavy_pairs = md.compute_contacts(
pdb, contacts, scheme='sidechain-heavy')
ref_ca = np.loadtxt(get_fn('cc_ca.dat'))
ref_closest = np.loadtxt(get_fn('cc_closest.dat'))
ref_closest_heavy = np.loadtxt(get_fn('cc_closest-heavy.dat'))
ref_sidechain = np.loadtxt(get_fn('cc_sidechain.dat'))
ref_sidechain_heavy = np.loadtxt(get_fn('cc_sidechain-heavy.dat'))
eq(ref_ca, ca.flatten())
eq(ref_closest, closest.flatten())
eq(ref_closest_heavy, closest_heavy.flatten())
eq(ref_sidechain, sidechain.flatten())
eq(ref_sidechain_heavy, sidechain_heavy.flatten())
eq(contacts, ca_pairs)
eq(contacts, closest_pairs)
eq(contacts, closest_heavy_pairs)
eq(contacts, sidechain_pairs)
eq(contacts, sidechain_heavy_pairs)
def shukla_coords(trajectories, KER, Aloop, SRC2):
difference = []
rmsd = []
for traj in trajectories:
# append difference
k295e310 = md.compute_contacts(traj, [KER[0]])
e310r409 = md.compute_contacts(traj, [KER[1]])
difference.append(
10 * (e310r409[0] -
k295e310[0])) # 10x because mdtraj is naturally in nm
# append rmsd
Activation_Loop_SRC2 = SRC2.top.select(
"backbone and (resid %s to %s)" % (Aloop[0], Aloop[1]))
Activation_Loop_kinase = traj.top.select(
"backbone and (resid %s to %s)" % (Aloop[0], Aloop[1]))
SRC2_cut = SRC2.atom_slice(Activation_Loop_SRC2)
traj_cut = traj.atom_slice(Activation_Loop_kinase)
rmsd.append(10 *
(md.rmsd(traj_cut, SRC2_cut,
frame=0))) # 10x because mdtraj is naturaly in nm
return [rmsd, difference]
def test_Residue_Mindist_Ca_array_periodic(self):
traj = mdtraj.load(pdbfile)
# Atoms most far appart in Z
atom_minz = traj.xyz.argmin(1).squeeze()[-1]
atom_maxz = traj.xyz.argmax(1).squeeze()[-1]
# Residues with the atoms most far appart in Z
res_minz = traj.topology.atom(atom_minz).residue.index
res_maxz = traj.topology.atom(atom_maxz).residue.index
contacts = np.array([[res_minz, res_maxz]])
# Tweak the trajectory so that a (bogus) PBC exists (otherwise traj._have_unitcell is False)
traj.unitcell_angles = [90, 90, 90]
traj.unitcell_lengths = [1, 1, 1]
self.feat.add_residue_mindist(scheme='ca',
residue_pairs=contacts,
periodic=False)
D = self.feat.transform(traj)
Dperiodic_true = mdtraj.compute_contacts(traj,
scheme='ca',
contacts=contacts,
periodic=True)[0]
Dperiodic_false = mdtraj.compute_contacts(traj,
scheme='ca',
contacts=contacts,
periodic=False)[0]
# This asserts that the periodic option is having an effect at all
assert not np.allclose(
Dperiodic_false,
Dperiodic_true,
)
# This asserts that the periodic option is being handled correctly by pyemma
assert np.allclose(D, Dperiodic_false)
assert len(self.feat.describe()) == self.feat.dimension()
def catkhrd(trajectories):
# define empty lists
D218 = []
D222 = []
for traj in trajectories:
#append h188s218 difference
h188s218 = md.compute_contacts(traj, [[120,151]],scheme='ca')
D218.append(h188s218[0])
#append k97s222 difference
k97s222 = md.compute_contacts(traj, [[29,155]],scheme='ca')
D222.append(k97s222[0])
#flatten these lists of arrays
flattened_h188s218 = np.asarray([val for sublist in D218 for val in sublist])
flattened_k97s222 = np.asarray([val for sublist in D222 for val in sublist])
return [flattened_h188s218, flattened_k97s222]
def shukla_coords_byrun(files,KER,Aloop,SRC2):
difference = []
rmsd = []
difference_combinetrajs = []
rmsd_combinetrajs = []
path_base = files.split('*')[0]
clone0_files = "%s/*clone0.h5" % path_base
globfiles = glob(clone0_files)
runs_list = []
for filename in globfiles:
run_string = re.search('run([^-]+)',filename).group(1)
run = int(run_string)
if run not in runs_list:
runs_list.append(run)
runs_list.sort()
for run in runs_list:
trajectories = dataset.MDTrajDataset("%s/run%d-clone*1.h5" % (path_base,run))
print "Run %s has %s trajectories." % (run,len(trajectories))
for traj in trajectories:
# append difference
k295e310 = md.compute_contacts(traj, [KER[0]])
e310r409 = md.compute_contacts(traj, [KER[1]])
difference_combinetrajs.append(10*(e310r409[0] - k295e310[0])) # 10x because mdtraj is naturally in nm
# append rmsd
Activation_Loop_SRC2 = SRC2.top.select("backbone and (resid %s to %s)" %(Aloop[0],Aloop[1]))
Activation_Loop_kinase = traj.top.select("backbone and (resid %s to %s)" %(Aloop[0],Aloop[1]))
SRC2_cut = SRC2.atom_slice(Activation_Loop_SRC2)
traj_cut = traj.atom_slice(Activation_Loop_kinase)
rmsd_combinetrajs.append(10*(md.rmsd(traj_cut,SRC2_cut,frame=0))) # 10x because mdtraj is naturaly in nm
# flatten list of arrays
difference_combinetrajs = np.asarray([val for sublist in difference_combinetrajs for val in sublist])
rmsd_combinetrajs = np.asarray([val for sublist in rmsd_combinetrajs for val in sublist])
difference.append(difference_combinetrajs)
difference_combinetrajs = []
rmsd.append(rmsd_combinetrajs)
rmsd_combinetrajs = []
return [rmsd, difference]
def read_and_featurize(traj_file, features_dir = None, condition=None, dihedral_types = ["phi", "psi", "chi1", "chi2"], dihedral_residues = None, resSeq_pairs = None, iterative = True):
a = time.time()
dihedral_indices = []
residue_order = []
if len(dihedral_residues) > 0:
for dihedral_type in dihedral_types:
if dihedral_type == "phi": dihedral_indices.append(phi_indices(fix_topology(top), dihedral_residues))
if dihedral_type == "psi": dihedral_indices.append(psi_indices(fix_topology(top), dihedral_residues))
if dihedral_type == "chi1": dihedral_indices.append(chi1_indices(fix_topology(top), dihedral_residues))
if dihedral_type == "chi2": dihedral_indices.append(chi2_indices(fix_topology(top), dihedral_residues))
#print("new features has dim %d" %(2*len(phi_tuples) + 2*len(psi_tuples) + 2*len(chi2_tuples)))
#print("feauturizing manually:")
dihedral_angles = []
for dihedral_type in dihedral_indices:
angles = np.transpose(ManualDihedral.compute_dihedrals(traj=traj,indices=dihedral_type))
dihedral_angles.append(np.sin(angles))
dihedral_angles.append(np.cos(angles))
manual_features = np.transpose(np.concatenate(dihedral_angles))
if len(resSeq_pairs) > 0:
top = md.load_frame(traj_file, index=0).topology
resIndex_pairs = convert_resSeq_to_resIndex(top, resSeq_pairs)
contact_features = []
if iterative:
try:
for chunk in md.iterload(traj_file, chunk = 1000):
# chunk = fix_traj(chunk)
#chunk = md.load(traj_file,stride=1000)
#print(resIndex_pairs[0:10])
chunk_features = md.compute_contacts(chunk, contacts = resIndex_pairs, scheme = 'closest-heavy', ignore_nonprotein=False)[0]
print(np.shape(chunk_features))
contact_features.append(chunk_features)
contact_features = np.concatenate(contact_features)
except Exception,e:
print str(e)
print("Failed")
return
#traj = md.load(traj_file)
#contact_features = md.compute_contacts(chunk, contacts = contact_residue_pairs, scheme = 'closest-heavy', ignore_nonprotein=False)[0]
else:
try:
traj = md.load(traj_file)
contact_features = md.compute_contacts(traj, contacts = resIndex_pairs, scheme = 'closest-heavy', ignore_nonprotein=False)[0]
except Exception,e:
print str(e)
print("Failed for traj")
return
def __init__(self,
native,
group1,
group2,
ca_cutoff_angstroms=10.,
verbose=True):
self.native = native[0] # ensure only a single frame is passed
res_group1, res_group2 = [
np.array(
sorted(set([native.topology.atom(i).residue.index
for i in g]))) for g in (group1, group2)
]
contact_pairs = np.array([(i, j) for i in res_group1
for j in res_group2])
is_contact = (
10. *
md.compute_contacts(native, scheme='ca', contacts=contact_pairs)[0]
< ca_cutoff_angstroms)[0]
contacts = contact_pairs[is_contact]
interface_residues = sorted(
set(contacts[:, 0]).union(set(contacts[:, 1])))
if verbose:
print '%i interface residues (%i,%i)' % (len(interface_residues),
len(set(contacts[:, 0])),
len(set(contacts[:, 1])))
self.interface_atom_indices = np.array([
a.index for a in native.topology.atoms
if a.residue.index in interface_residues
])
def get_interface_contacts(frame, ca_cutoff_ang=10.):
"""
Identify interface residues between ligand chains and receptor chains using mdtraj.
Residues identified by user-specified c-alpha cutoff, preset to 10 angstroms.
Feeds into contacts_iterator.
"""
#Get list of residues in receptor and ligand
r_residues = []
for chain in self.receptor_chains:
r_residues.extend([residue.index for residue in frame.topology.chain(chain).residues])
l_residues = []
for chain in self.ligand_chains:
l_residues.extend([residue.index for residue in frame.topology.chain(chain).residues])
# Make an array of potential contact pairs between receptor and ligand
contact_pairs = np.array([(i,j) for i in r_residues for j in l_residues])
# Check which ones fall within c-alpha distance cutoff
is_contact = (10.*md.compute_contacts(frame, scheme='ca', contacts=contact_pairs)[0] < ca_cutoff_ang)[0]
# Go from bool truth values to the actual residues
contacts = contact_pairs[is_contact]
# Go from pairs to flattened list of unique residues involved in contacts
self.interface_residues = sorted(set(contacts[:,0]).union(set(contacts[:,1])))
return self.interface_residues
def __init__(self, struct):
# read the structure from a pdb file (one chain, no hetatm, no water, just the good old protein, please
self.struct = struct
# compute the contacts. This computes all the minimum atom distances between residues. The cutoff will be applied later
self.dd, self.rp = mdtraj.compute_contacts(self.struct)
# ah si, this is the sequence of the pdb chain (the residues which are resolved, may be less than the FASTA from DB website)
self.seq = ''.join([r.code for r in self.struct.topology.residues])
def maker_w_strings(data):
i, sdf_state, num_residues, pdb = data
tmp_file = str(id(multiprocessing.current_process()))
create_pdb(sdf_state, pdb, tmp_file) # create the pdb for one state
# Load the file into mdtraj
t = md.load("tmp{0}.pdb".format(tmp_file))
pl = len(list(t.topology.residues))
resi = np.arange(pl)
pairs = list(itertools.product(resi, resi))
# compute constacts
matrix, l = md.compute_contacts(t, contacts=pairs, scheme="closest-heavy")
matrix = np.array(matrix).reshape((pl, pl))
### generate protein-ligand mask
mask = np.zeros((pl, pl))
for i in range(pl - num_residues):
for j in range(pl - num_residues):
# ligand
mask[i, j] = -1
for i in range(pl - num_residues, pl):
for j in range(pl - num_residues, pl):
# protein
mask[i, j] = 1
p_img = cv2.resize(matrix, dsize=(64, 64),
interpolation=cv2.INTER_CUBIC).reshape(64, 64, 1)
m_img = cv2.resize(mask, dsize=(64, 64),
interpolation=cv2.INTER_CUBIC).reshape(64, 64, 1)
p_contact_matrix = np.concatenate([p_img, m_img], axis=-1)
os.remove("tmp{0}.pdb".format(tmp_file))
return p_contact_matrix
def get_dists(filename):
conf = md.load(filename)
num_peptide_residues = len(conf.top.select("name == CA and chainid == 2"))
num_binding_site_residues = 180
num_total_residues = len(conf.top.select("name == CA"))
peptide_indices = range(num_total_residues)[-num_peptide_residues:]
binding_site_indices = range(num_binding_site_residues)
contacts_all = list(
itertools.product(peptide_indices, binding_site_indices))
#print("Num contacts:", len(contacts_all))
distances, pairs = md.compute_contacts(conf, contacts_all)
distances = distances[0] # only a single frame
interactions = []
resres_dists = []
residues = [r for r in conf.top.residues]
for i, p in enumerate(pairs):
r1 = str(residues[p[0]])[:3]
r2 = str(residues[p[1]])[:3]
interactions.append(r1 + "-" + r2)
resres_dists.append(distances[i] * 10)
return interactions, resres_dists
def featurize_sig(filename):
conf = md.load(filename)
num_peptide_residues = len(conf.top.select("name == CA and chainid == 2"))
num_binding_site_residues = 180
num_total_residues = len(conf.top.select("name == CA"))
peptide_indices = range(num_total_residues)[-num_peptide_residues:]
binding_site_indices = range(num_binding_site_residues)
contacts_all = list(
itertools.product(peptide_indices, binding_site_indices))
#print("Num contacts:", len(contacts_all))
distances, pairs = md.compute_contacts(conf, contacts_all)
distances = distances[0] # only a single frame
feature_vec = np.zeros((num_dim, ))
residues = [r for r in conf.top.residues]
for i, p in enumerate(pairs):
r1 = str(residues[p[0]])[:3]
r2 = str(residues[p[1]])[:3]
resres_dist = distances[i] * 10
alpha = 5 #np.log(99)+4
feature_vec[new_interaction_to_index[
r1 + "-" + r2]] += 1. / (1 + np.exp(resres_dist - alpha))
if abs(distances[i]) < 0.001: print(r1, r2, distances[i], p)
return feature_vec
def get_distances(filename):
conf = md.load(filename)
num_peptide_residues = len(conf.top.select("name == CA and chainid == 2"))
num_binding_site_residues = 180
num_total_residues = len(conf.top.select("name == CA"))
peptide_indices = range(num_total_residues)[-num_peptide_residues:]
binding_site_indices = range(num_binding_site_residues)
contacts_all = list(
itertools.product(peptide_indices, binding_site_indices))
#print("Num contacts:", len(contacts_all))
distances, pairs = md.compute_contacts(conf, contacts_all)
distances = distances[0] # only a single frame
feature_vec = np.zeros((num_dim, ))
residues = [r for r in conf.top.residues]
resres_names = []
pep_mhc_distances = []
for i, p in enumerate(pairs):
r1 = str(residues[p[0]])
r2 = str(residues[p[1]])
resres_names.append([r1, r2])
pep_mhc_distances.append(distances[i] * 10)
#feature_vec[new_interaction_to_index[r1+"-"+r2]] += 1./(distances[i]*10)
#if abs(distances[i]) < 0.001: print(r1, r2, distances[i], p)
return resres_names, pep_mhc_distances
def compute_contacts_below_cutoff(traj_file_frame, cutoff = 100000.0, contact_residues = [], anton = False):
traj_file = traj_file_frame[0]
frame = md.load_frame(traj_file, index = 0)
#frame = fix_traj(frame)
top = frame.topology
distance_residues = []
res_indices = []
resSeq_to_resIndex = {}
residue_full_infos = []
for i in range(0, len(contact_residues)):
residue = contact_residues[i]
indices = [r.index for r in top.residues if r.resSeq == residue[1] and r.chainid == residue[0] and not r.is_water]
if len(indices) == 0:
print("No residues in trajectory for residue %d" %residue)
continue
else:
ind = indices[0]
for j in indices:
if j != ind:
#print("Warning: multiple res objects for residue %d " %residue)
if "CB" in [str(a) for a in r.atoms for r in top.residues if r.index == ind]:
ind = j
res_indices.append(ind)
distance_residues.append(residue)
resSeq_to_resIndex[residue] = ind
resSeq_combinations = itertools.combinations(distance_residues, 2)
res_index_combinations = []
resSeq_pairs = [c for c in resSeq_combinations]
for combination in resSeq_pairs:
res0 = combination[0]
res1 = combination[1]
res_index0 = resSeq_to_resIndex[res0]
res_index1 = resSeq_to_resIndex[res1]
res_index_combinations.append((res_index0, res_index1))
final_resSeq_pairs = []
final_resIndex_pairs = []
distances = md.compute_contacts(frame, contacts = res_index_combinations, scheme = 'closest-heavy', ignore_nonprotein=False)[0]
#print(distances)
print(np.shape(distances))
for i in range(0, len(distances[0])):
distance = distances[0][i]
#print(distance)
if distance < cutoff:
final_resIndex_pairs.append(res_index_combinations[i])
final_resSeq_pairs.append(resSeq_pairs[i])
for pair in final_resIndex_pairs:
info0 = [(r.resSeq, r.name, r.chain.index) for r in top.residues if r.index == pair[0]]
info1 = [(r.resSeq, r.name, r.chain.index) for r in top.residues if r.index == pair[1]]
residue_full_infos.append((info0, info1))
print(len(final_resSeq_pairs))
print(len(final_resIndex_pairs))
return((final_resSeq_pairs, residue_full_infos))
def partial_transform(self, traj):
"""Featurize an MD trajectory into a vector space derived from
residue-residue distances
Parameters
----------
traj : mdtraj.Trajectory
A molecular dynamics trajectory to featurize.
Returns
-------
features : np.ndarray, dtype=float, shape=(n_samples, n_features)
A featurized trajectory is a 2D array of shape
`(length_of_trajectory x n_features)` where each `features[i]`
vector is computed by applying the featurization function
to the `i`th snapshot of the input trajectory.
See Also
--------
transform : simultaneously featurize a collection of MD trajectories
"""
distances, _ = md.compute_contacts(traj, self.contacts,
self.scheme, self.ignore_nonprotein)
return self._transform(distances)
def igeom2mindist_COMdist_truncation(igeom,
res_COM_cutoff_Ang=25,
):
COMs_xyz = geom2COMxyz(igeom)
COMs_dist_triu = _np.array([pdist(ixyz) for ixyz in COMs_xyz])
COMs_under_cutoff = COM_n_from_COM_dist_triu(COMs_dist_triu,
cutoff_nm=res_COM_cutoff_Ang/10)
COMs_under_cutoff_pair_idxs = _np.argwhere(COMs_under_cutoff.sum(0) >= 1).squeeze()
pairs = _np.vstack(_np.triu_indices(igeom.n_residues, 1)).T[COMs_under_cutoff_pair_idxs]
try:
ctcs, ctc_idxs_dummy = _md.compute_contacts(igeom, pairs)
except MemoryError:
print("\nCould not fit %u contacts for %u frames into memory"%(len(pairs), igeom.n_frames))
raise
assert _np.allclose(pairs, ctc_idxs_dummy)
return ctcs.min(0), pairs, COMs_under_cutoff_pair_idxs
def describe_features(self, traj):
"""Return a list of dictionaries describing the features in Contacts."""
x = []
# fill in the atom indices using just the first frame
distances, residue_indices = md.compute_contacts(traj, self.contacts, self.scheme, self.ignore_nonprotein)
n = residue_indices.shape[0]
aind = ["N/A"] * n
resSeq = [np.array([traj.top.residue(j).resSeq for j in i]) for i in residue_indices]
resid = [np.array([traj.top.residue(j).index for j in i]) for i in residue_indices]
resnames = [[traj.topology.residue(j).name for j in i] for i in resid]
bigclass = [self.contacts] * n
smallclass = [self.scheme] * n
otherInfo = [self.ignore_nonprotein] * n
for i in range(n):
d_i = dict(
resname=resnames[i],
atomind=aind[i],
resSeq=resSeq[i],
resid=resid[i],
otherInfo=otherInfo[i],
bigclass=bigclass[i],
smallclass=smallclass[i],
)
x.append(d_i)
return x
def _Compute_Contacts_Between_Residues(self,
Pairs=None,
Cutoff=0.5,
Trajectory=None):
output = mdtraj.compute_contacts(Trajectory, Pairs)
distances = output[0].T
atom_pairs = output[1]
residues_in_contact = []
contacts = []
idx = 0
cont = 0
for frame in distances:
for d in frame:
if (d < Cutoff):
atom_pair = atom_pairs[idx]
resid1 = self.top.residue(atom_pair[0])
resid2 = self.top.residue(atom_pair[1])
arr = [resid1, resid2]
if (arr not in residues_in_contact):
residues_in_contact.append(arr)
cont += 1
contacts.append(cont)
cont = 0
idx += 1
return (residues_in_contact, contacts)
def partial_transform(self, traj):
"""Featurize an MD trajectory into a vector space derived from
residue-residue distances
Parameters
----------
traj : mdtraj.Trajectory
A molecular dynamics trajectory to featurize.
Returns
-------
features : np.ndarray, dtype=float, shape=(n_samples, n_features)
A featurized trajectory is a 2D array of shape
`(length_of_trajectory x n_features)` where each `features[i]`
vector is computed by applying the featurization function
to the `i`th snapshot of the input trajectory.
See Also
--------
transform : simultaneously featurize a collection of MD trajectories
"""
# check to make sure topologies are consistent with the reference frame
try:
assert traj.top == self.reference_frame.top
except:
warnings.warn("The topology of the trajectory is not" +
"the same as that of the reference frame," +
"which might give meaningless results.")
distances, _ = md.compute_contacts(traj, self.contacts,
self.scheme, ignore_nonprotein=False)
return self._transform(distances)
def test_contact_4(get_fn):
pdb = md.load(
get_fn('1am7_protein.pdb')
) # protonated and including at least one glycine residue (which has no heavy atoms in its sidechain)
contacts = md.compute_contacts(
pdb, contacts='all', scheme='sidechain-heavy'
) # test passes if this doesn't raise an exception
def test_Residue_Mindist_Ca_all(self):
n_ca = self.feat.topology.n_atoms
self.feat.add_residue_mindist(scheme='ca')
D = self.feat.transform(self.traj)
Dref = mdtraj.compute_contacts(self.traj, scheme='ca')[0]
assert np.allclose(D, Dref)
assert len(self.feat.describe()) == self.feat.dimension()
def test_Residue_Mindist_Ca_array(self):
contacts=np.array([[20,10,], [10,0]])
self.feat.add_residue_mindist(scheme='ca', residue_pairs=contacts)
D = self.feat.transform(self.traj)
Dref = mdtraj.compute_contacts(self.traj, scheme='ca', contacts=contacts)[0]
assert np.allclose(D, Dref)
assert len(self.feat.describe())==self.feat.dimension()
def partial_transform(self, traj):
"""Featurize an MD trajectory into a vector space derived from
residue-residue distances
Parameters
----------
traj : mdtraj.Trajectory
A molecular dynamics trajectory to featurize.
Returns
-------
features : np.ndarray, dtype=float, shape=(n_samples, n_features)
A featurized trajectory is a 2D array of shape
`(length_of_trajectory x n_features)` where each `features[i]`
vector is computed by applying the featurization function
to the `i`th snapshot of the input trajectory.
See Also
--------
transform : simultaneously featurize a collection of MD trajectories
"""
# check to make sure topologies are consistent with the reference frame
try:
assert traj.top == self.reference_frame.top
except:
warnings.warn("The topology of the trajectory is not" +
"the same as that of the reference frame," +
"which might give meaningless results.")
distances, _ = md.compute_contacts(traj, self.contacts,
self.scheme, ignore_nonprotein=False,
periodic = self.periodic)
return self._transform(distances)
def describe_features(self, traj):
"""Return a list of dictionaries describing the features in Contacts."""
x = []
# fill in the atom indices using just the first frame
distances, residue_indices = md.compute_contacts(
traj, self.contacts, self.scheme, self.ignore_nonprotein)
n = residue_indices.shape[0]
aind = ["N/A"] * n
resSeq = [
np.array([traj.top.residue(j).resSeq for j in i])
for i in residue_indices
]
resid = [
np.array([traj.top.residue(j).index for j in i])
for i in residue_indices
]
resnames = [[traj.topology.residue(j).name for j in i] for i in resid]
bigclass = [self.contacts] * n
smallclass = [self.scheme] * n
otherInfo = [self.ignore_nonprotein] * n
for i in range(n):
d_i = dict(resname=resnames[i],
atomind=aind[i],
resSeq=resSeq[i],
resid=resid[i],
otherInfo=otherInfo[i],
bigclass=bigclass[i],
smallclass=smallclass[i])
x.append(d_i)
return x
def compute_mdtraj_order_parmeters(trajectory_file, rmsd_reference_structure=None):
# documentation: http://mdtraj.org/1.8.0/analysis.html#
trajectory = md.load(trajectory_file)
return_values = []
return_value_names = []
if not rmsd_reference_structure == None:
reference = md.load(rmsd_reference_structure)
rmsd = md.rmsd(trajectory, reference)
return_values.append(rmsd)
return_value_names.append("RMSD")
hydrogen_bonds = np.array([np.sum(x) for x in md.kabsch_sander(trajectory)])
return_values.append(hydrogen_bonds)
return_value_names.append("HBondEnergy")
ss = md.compute_dssp(trajectory)
shape = ss.shape
transdict = dict(zip(list(set(list(ss.flatten()))),range(len(list(set(list(ss.flatten())))))))
ss = np.array([transdict[x] for x in ss.flatten()]).reshape(shape).T
return_values.append(ss)
return_value_names.append("SecondaryStructure")
rg = md.compute_rg(trajectory)
return_values.append(rg)
return_value_names.append("Rg")
distances, residue_pairs = md.compute_contacts(trajectory, scheme='ca')
contacts = md.geometry.squareform(distances, residue_pairs)
return_values.append(contacts)
return_value_names.append("Contacts")
return dict(zip(return_value_names, return_values))
def partial_transform(self, traj):
"""Featurize an MD trajectory into a vector space derived from
residue-residue distances
Parameters
----------
traj : mdtraj.Trajectory
A molecular dynamics trajectory to featurize.
Returns
-------
features : np.ndarray, dtype=float, shape=(n_samples, n_features)
A featurized trajectory is a 2D array of shape
`(length_of_trajectory x n_features)` where each `features[i]`
vector is computed by applying the featurization function
to the `i`th snapshot of the input trajectory.
See Also
--------
transform : simultaneously featurize a collection of MD trajectories
"""
distances, _ = md.compute_contacts(traj, self.contacts, self.scheme,
self.ignore_nonprotein)
return self._transform(distances)
def contacts_bonds(traj, peptide_chain ): group_1 = [residue.index for residue in traj.topology.chain(peptide_chain).residues ] group_2 = [residue.index for residue in traj.topology.chain(0).residues or traj.topology.chain(1).residues or traj.topology.chain(2).residues ] pairs = list(product(group_1, group_2)) contacts_bonds= mdtraj.compute_contacts(traj,pairs , scheme='closest-heavy', ignore_nonprotein=True, periodic=True, soft_min=False, soft_min_beta=20)
def plot_native_state_contact_map(title):
colors = [('white')] + [(cm.jet(i)) for i in xrange(1,256)]
new_map = matplotlib.colors.LinearSegmentedColormap.from_list('new_map', colors, N=256)
if os.path.exists("contact_pairs.dat") and os.path.exists("contact_probabilities.dat"):
pairs = np.loadtxt("contact_pairs.dat")
probability = np.loadtxt("contact_probabilities.dat")
else:
print " Loading BeadBead.dat"
beadbead = np.loadtxt("BeadBead.dat",dtype=str)
sigij = beadbead[:,5].astype(float)
epsij = beadbead[:,6].astype(float)
deltaij = beadbead[:,7].astype(float)
interaction_numbers = beadbead[:,4].astype(str)
pairs = beadbead[:,:2].astype(int)
pairs -= np.ones(pairs.shape,int)
np.savetxt("contact_pairs.dat",pairs)
print " Computing distances with mdtraj..."
traj = md.load("traj.xtc",top="Native.pdb")
distances = md.compute_contacts(traj,pairs)
contacts = (distances[0][:] <= 1.2*sigij).astype(int)
print " Computing contact probability..."
probability = sum(contacts.astype(float))/contacts.shape[0]
np.savetxt("contact_probabilities.dat",probability)
Qref = np.loadtxt("Qref_cryst.dat")
C = np.zeros(Qref.shape,float)
for k in range(len(pairs)):
C[pairs[k][0],pairs[k][1]] = probability[k]
print " Plotting..."
plt.figure()
plt.subplot(1,1,1,aspect=1)
ax = plt.subplot(1,1,1,aspect=1)
plt.pcolor(C,cmap=new_map)
for k in range(len(pairs)):
if probability[k] > 0.01:
plt.plot(pairs[k][1],pairs[k][0],marker='s',ms=3.0,markeredgecolor=new_map(probability[k]),color=new_map(probability[k]))
else:
continue
plt.xlim(0,len(Qref))
plt.ylim(0,len(Qref))
#plt.text(10,70,name.upper(),fontsize=70,color="r")
ax = plt.gca()
cbar = plt.colorbar()
cbar.set_clim(0,1)
cbar.set_label("Contact probability",fontsize=20)
cbar.ax.tick_params(labelsize=20)
plt.xlabel("Residue i",fontsize=20)
plt.ylabel("Residue j",fontsize=20)
#plt.title("Native State Contact Map "+title,fontsize=20)
plt.title(title)
for label in ax.get_xticklabels() + ax.get_yticklabels():
label.set_fontsize(15)
print " Saving..."
plt.savefig("native_state_contact_map.pdf")
def ca_contact_pca(traj, n_pc, cutoff_angstroms=8., variance_scaled=True):
from sklearn.decomposition import TruncatedSVD
m = (10.*md.compute_contacts(traj,scheme='ca')[0]<cutoff_angstroms)
m = m-m.mean(axis=0)
trunc_svd = TruncatedSVD(n_pc).fit(m)
pc = trunc_svd.transform(m)*(trunc_svd.explained_variance_ratio_ if variance_scaled else 1.)
del m
return pc
def compute_contacts(self, residue_pairs):
"""
:param residue_pairs: An array containing pairs of indices (0-indexed) of residues to compute the contacts between
:return: distances: np.ndarray, shape=(n_frames, n_pairs); residues_pairs: np.ndarray, shape=(n_pairs, 2)
"""
return md.compute_contacts(self.traj, residue_pairs)
def test_contact_3(get_fn):
pdb = md.load(get_fn('bpti.pdb'))
beta = 20
dists, pairs = md.compute_contacts(pdb, soft_min=True, soft_min_beta=beta)
maps = md.geometry.squareform(dists, pairs)
for i, (r0, r1) in enumerate(pairs):
for t in range(pdb.n_frames):
assert np.allclose(beta / np.log(np.sum(np.exp(beta / maps[t, r0, r1]))), dists[t, i])
def _distances(traj: md.Trajectory, scheme: str, transform: str,
centre: Union[float, None], steepness: Union[float, None]):
feat, ix = md.compute_contacts(traj, contacts='all', scheme=scheme)
if transform == 'logistic':
assert (centre is not None) and (steepness is not None)
tmp = 1.0/(1.+np.exp((-1)*steepness*(feat-centre)))
assert np.allclose(tmp.shape, feat.shape)
feat = tmp
return feat
def _get_contact_pairs(self, contacts):
if self.scheme == 'ca':
if not any(
a
for a in self.reference_frame.top.chain(ligand_chain).atoms
if a.name.lower() == 'ca'):
raise ValueError("Bad scheme: the ligand has no alpha carbons")
# this is really similar to mdtraj/contact.py, but ensures that
# md.compute_contacts is always seeing an array of exactly the
# contacts we want to specify
if isinstance(contacts, string_types):
if contacts.lower() != 'all':
raise ValueError(
'({}) is not a valid contacts specifier'.format(
contacts.lower()))
self.residue_pairs = []
for i in np.arange(
self.reference_frame.top.chain(
self.protein_chain).n_residues):
for j in np.arange(
self.reference_frame.top.chain(
self.ligand_chain).n_residues):
self.residue_pairs.append(
(i + self.p_residue_offset, j + self.l_residue_offset))
self.residue_pairs = np.array(self.residue_pairs)
if len(self.residue_pairs) == 0:
raise ValueError('No acceptable residue pairs found')
else:
self.residue_pairs = ensure_type(np.asarray(contacts),
dtype=np.int,
ndim=2,
name='contacts',
shape=(None, 2),
warn_on_cast=False)
if not np.all(
(self.residue_pairs >= 0) *
(self.residue_pairs < self.reference_frame.n_residues)):
raise ValueError('contacts requests a residue that is not '\
'in the permitted range')
if self.binding_pocket is not 'all':
ref_distances, _ = md.compute_contacts(self.reference_frame,
self.residue_pairs,
self.scheme,
ignore_nonprotein=False)
self.residue_pairs = self.residue_pairs[np.where(
ref_distances < self.binding_pocket)[1]]
if len(self.residue_pairs) == 0:
raise ValueError('No residue pairs within binding pocket')
return self.residue_pairs
def test_Residue_Mindist_Ca_all_threshold(self):
threshold = .7
self.feat.add_residue_mindist(scheme='ca', threshold=threshold)
D = self.feat.transform(self.traj)
Dref = mdtraj.compute_contacts(self.traj, scheme='ca')[0]
Dbinary = np.zeros_like(Dref)
I = np.argwhere(Dref <= threshold)
Dbinary[I[:, 0], I[:, 1]] = 1
assert np.allclose(D, Dbinary)
assert len(self.feat.describe()) == self.feat.dimension()
def get_pocket_residues(traj):
traj = traj.atom_slice(traj.topology.select("protein or resn UNL"))
resn = len(list(traj.topology.residues))
group_1 = list(range(resn - 1))
group_2 = [resn - 1]
pairs = list(itertools.product(group_1, group_2))
res, pairs = md.compute_contacts(traj, pairs)
pocket_resids = list(np.where(res[0] <= 0.5)[0] + 1)
pocket_resids = ["resid {}".format(id) for id in pocket_resids]
pocket_resids = " or ".join(pocket_resids)
return pocket_resids
def shukla_coords(trajectories,KER,Aloop,SRC2):
difference = []
rmsd = []
for traj in trajectories:
# append difference
k295e310 = md.compute_contacts(traj, [KER[0]])
e310r409 = md.compute_contacts(traj, [KER[1]])
difference.append(10*(e310r409[0] - k295e310[0])) # 10x because mdtraj is naturally in nm
# append rmsd
Activation_Loop_SRC2 = SRC2.top.select("backbone and (resid %s to %s)" %(Aloop[0],Aloop[1]))
Activation_Loop_kinase = traj.top.select("backbone and (resid %s to %s)" %(Aloop[0],Aloop[1]))
SRC2_cut = SRC2.atom_slice(Activation_Loop_SRC2)
traj_cut = traj.atom_slice(Activation_Loop_kinase)
rmsd.append(10*(md.rmsd(traj_cut,SRC2_cut,frame=0))) # 10x because mdtraj is naturaly in nm
return [rmsd, difference]
def test_contact_1():
pdb = md.load(get_fn('bpti.pdb'))
dists, pairs = md.compute_contacts(pdb)
for r0, r1 in pairs:
# are these valid residue indices?
pdb.topology.residue(r0)
pdb.topology.residue(r1)
assert not (abs(r0 - r1) < 3)
maps = md.geometry.squareform(dists, pairs)
for i, (r0, r1) in enumerate(pairs):
for t in range(pdb.n_frames):
eq(maps[t, r0, r1], dists[t, i])
def test_ContactFeaturizer_describe_features():
scheme = np.random.choice(['ca','closest','closest-heavy'])
feat = ContactFeaturizer(scheme=scheme, ignore_nonprotein=True)
rnd_traj = np.random.randint(len(trajectories))
features = feat.transform([trajectories[rnd_traj]])
df = pd.DataFrame(feat.describe_features(trajectories[rnd_traj]))
for f in range(25):
f_index = np.random.choice(len(df))
residue_ind = df.iloc[f_index].resids
feature_value, _ = md.compute_contacts(trajectories[rnd_traj],
contacts=[residue_ind],
scheme=scheme)
assert (features[0][:, f_index] == feature_value.flatten()).all()
def describe_features(self, traj):
"""Return a list of dictionaries describing the contacts features.
Parameters
----------
traj : mdtraj.Trajectory
The trajectory to describe
Returns
-------
feature_descs : list of dict
Dictionary describing each feature with the following information
about the atoms participating in each dihedral
- resnames: unique names of residues
- atominds: the four atom indicies
- resseqs: unique residue sequence ids (not necessarily
0-indexed)
- resids: unique residue ids (0-indexed)
- featurizer: Contact
- featuregroup: ca, heavy etc.
"""
feature_descs = []
# fill in the atom indices using just the first frame
distances, residue_indices = md.compute_contacts(traj[0],
self.contacts, self.scheme,
ignore_nonprotein=False,
periodic=self.periodic)
top = traj.topology
aind = []
resseqs = []
resnames = []
for resid_ids in residue_indices:
aind += ["N/A"]
resseqs += [[top.residue(ri).resSeq for ri in resid_ids]]
resnames += [[top.residue(ri).name for ri in resid_ids]]
zippy = itertools.product(["Ligand Contact"], [self.scheme],
["N/A"],
zip(aind, resseqs, residue_indices, resnames))
feature_descs.extend(dict_maker(zippy))
return feature_descs
def _get_contact_pairs(self, contacts):
if self.scheme=='ca':
if not any(a for a in self.reference_frame.top.chain(self.ligand_chain).atoms
if a.name.lower() == 'ca'):
raise ValueError("Bad scheme: the ligand has no alpha carbons")
# this is really similar to mdtraj/contact.py, but ensures that
# md.compute_contacts is always seeing an array of exactly the
# contacts we want to specify
if isinstance(contacts, string_types):
if contacts.lower() != 'all':
raise ValueError('({}) is not a valid contacts specifier'.format(contacts.lower()))
self.residue_pairs = []
for i in np.arange(self.reference_frame.top.chain(self.protein_chain).n_residues):
for j in np.arange(self.reference_frame.top.chain(self.ligand_chain).n_residues):
self.residue_pairs.append((i+self.p_residue_offset,
j+self.l_residue_offset))
self.residue_pairs = np.array(self.residue_pairs)
if len(self.residue_pairs) == 0:
raise ValueError('No acceptable residue pairs found')
else:
self.residue_pairs = ensure_type(np.asarray(contacts),
dtype=np.int, ndim=2, name='contacts',
shape=(None, 2), warn_on_cast=False)
if not np.all((self.residue_pairs >= 0) *
(self.residue_pairs < self.reference_frame.n_residues)):
raise ValueError('contacts requests a residue that is not '\
'in the permitted range')
if self.binding_pocket is not 'all':
ref_distances, _ = md.compute_contacts(self.reference_frame,
self.residue_pairs, self.scheme,
ignore_nonprotein=False, periodic = self.periodic)
self.residue_pairs = self.residue_pairs[np.where(ref_distances<
self.binding_pocket)[1]]
if len(self.residue_pairs) == 0:
raise ValueError('No residue pairs within binding pocket')
return self.residue_pairs
def test_contact_2():
pdb = md.load(get_fn('1vii_sustiva_water.pdb'))
dists, pairs = md.compute_contacts(pdb, scheme='closest')
for r0, r1 in pairs:
assert pdb.topology.residue(r0).name != 'HOH'
assert pdb.topology.residue(r1).name != 'HOH'
# spot check one of the pairs
r0, r1 = pairs[10]
atoms_r0 = [a.index for a in pdb.topology.residue(r0).atoms]
atoms_r1 = [a.index for a in pdb.topology.residue(r1).atoms]
atomdist = md.compute_distances(pdb, list(itertools.product(atoms_r0, atoms_r1)))
np.testing.assert_array_equal(dists[:, 10], np.min(atomdist, axis=1))
maps = md.geometry.squareform(dists, pairs)
for i, (r0, r1) in enumerate(pairs):
for t in range(pdb.n_frames):
eq(maps[t, r0, r1], dists[t, i])
def find_respairs_that_changed(fnames,
scheme = 'ca', # or 'closest' or 'closest-heavy'
threshold = 0.4,
stride = 100,
max_respairs = 1000):
'''
Parameters
----------
fnames : list of paths to trajectories
scheme : 'ca' or 'closest' or 'closest-heavy'
threshold : float
contact threshold (nm)
'''
distances = []
for fname in fnames:
traj = md.load(fname,stride=stride)
pairwise_distances,residue_pairs = md.compute_contacts(traj,scheme=scheme)
distances.append(pairwise_distances)
distances = np.vstack(distances)
# identify contacts that change by counting how many times the distances were
# greater than and less than the threshold
num_times_greater_than = (distances>threshold).sum(0)
num_times_less_than = (distances<threshold).sum(0)
changed = (num_times_greater_than > 0) * (num_times_less_than > 0)
print("Number of contacts that changed: {0}".format(changed.sum()))
print("Total number of possible contacts: {0}".format(len(residue_pairs)))
if len(changed) > max_respairs:
n_diff = np.min(np.vstack((num_times_less_than,num_times_greater_than)),0)
indices = sorted(np.arange(len(n_diff)),key=lambda i:-n_diff[i])
changed = indices[:max_respairs]
# now turn this bitmask into a list of relevant residue pairs
respairs_that_changed = residue_pairs[changed]
return respairs_that_changed
def prepare_trajectory(self, trajectory):
"""Prepare a trajectory for distance calculations based on the contact map.
Each frame in the trajectory will be represented by a vector where
each entries represents the distance between two residues in the structure.
Depending on what contacts you pick to use, this can be a 'native biased'
picture or not.
Paramters
---------
trajectory : mdtraj.Trajectory
The trajectory to prepare
Returns
-------
pairwise_distances : ndarray
1D array of various residue-residue distances
"""
# the result of md.compute_contacts is a tuple, where the distances are
# returned in the first element, and a list of contacts calculated are
# returned in the second element
return md.compute_contacts(trajectory, self.contacts, self.scheme)[0]
def get_distances(fname, scheme, stride):
'''
Function callable by a multiprocessing Pool
Parameters
----------
fname : string
filename of trajectory
scheme : string
'ca' or 'closest' or 'closest-heavy'
stride : int
thinning factor: only look at every `stride`th frame
Returns
-------
pairwise_distances : numpy array
residue_pairs : list of tuples
'''
traj = md.load(fname, stride = stride)
pairwise_distances,residue_pairs = md.compute_contacts(traj, scheme = scheme)
return pairwise_distances, residue_pairs
def read_and_featurize_iter(traj_file, features_dir = None, condition=None, dihedral_types = ["phi", "psi", "chi1", "chi2"], dihedral_residues = None, contact_residues = None):
a = time.time()
dihedral_indices = []
residue_order = []
if len(dihedral_residues) > 0:
for dihedral_type in dihedral_types:
if dihedral_type == "phi": dihedral_indices.append(phi_indices(fix_topology(top), dihedral_residues))
if dihedral_type == "psi": dihedral_indices.append(psi_indices(fix_topology(top), dihedral_residues))
if dihedral_type == "chi1": dihedral_indices.append(chi1_indices(fix_topology(top), dihedral_residues))
if dihedral_type == "chi2": dihedral_indices.append(chi2_indices(fix_topology(top), dihedral_residues))
#print("new features has dim %d" %(2*len(phi_tuples) + 2*len(psi_tuples) + 2*len(chi2_tuples)))
#print("feauturizing manually:")
dihedral_angles = []
for dihedral_type in dihedral_indices:
angles = np.transpose(ManualDihedral.compute_dihedrals(traj=traj,indices=dihedral_type))
dihedral_angles.append(np.sin(angles))
dihedral_angles.append(np.cos(angles))
manual_features = np.transpose(np.concatenate(dihedral_angles))
if len(contact_residues) > 0:
contact_features = []
for chunk in md.iterload(traj_file, chunk = 10000):
fixed_traj = fix_traj(chunk)
fixed_top = fixed_traj.topology
distance_residues = []
res_objects = [r for r in fixed_top.residues]
for r in contact_residues:
for res in res_objects:
if res.resSeq == r and len(res._atoms) > 5:
#print res._atoms
distance_residues.append(res.index)
if len(contact_residues) != len(distance_residues):
print "Residues are missing"
print len(contact_residues)
print len(distance_residues)
#sys.exit()
#return None
combinations = itertools.combinations(distance_residues, 2)
pairs = [c for c in combinations]
#print pairs
contact_features.append(md.compute_contacts(fixed_traj, contacts = pairs, scheme = 'closest-heavy', ignore_nonprotein=False)[0])
contact_features = np.concatenate(contact_features)
if len(dihedral_residues) > 0:
manual_features = np.column_stack((manual_features, contact_features))
else:
manual_features = contact_features
b = time.time()
print("new features %s has shape: " %traj_file)
print(np.shape(manual_features))
if condition is None:
condition = get_condition(traj_file)
verbosedump(manual_features, "%s/%s.h5" %(features_dir, condition))
def featurize(path_to_files,model_name):
files = glob(path_to_files)
print("Number of files: {0}".format(len(files)))
# timestep between frames: 250 picoseconds
#### A. FEATURE EXTRACTION ####
### Step 1: identifying interresidue contacts that change
# compute full contact maps for a strided subset of the simulation frames
strided_distances=[]
stride=100 # stride within trajectory
traj_thin=5 # only look at 1 in traj_thin trajectories
scheme = 'ca'
threshold = 0.8 # contact threshold in angstroms
for f in files[::traj_thin]:
traj = md.load(f,stride=stride)
distances,residue_pairs = md.compute_contacts(traj,scheme=scheme)
strided_distances.append(distances)
strided_distances = np.vstack(strided_distances)
# identify contacts that change by counting how many times the distances were
# greater than and less than the threshold
num_times_greater_than = (strided_distances>threshold).sum(0)
num_times_less_than = (strided_distances<threshold).sum(0)
changed = (num_times_greater_than > 0) * (num_times_less_than > 0)
print("Number of contacts that changed: {0}".format(changed.sum()))
print("Total number of possible contacts: {0}".format(len(residue_pairs)))
# now turn this bitmask into a list of relevant residue pairs
respairs_that_changed = residue_pairs[changed]
# save this list!
np.save('{0}_respairs_that_changed.npy'.format(model_name),respairs_that_changed)
### Step 2: extract these selected features from the full dataset
X = []
traj_thin=1 # only look at 1 in traj_thin trajectories
files_of_interest = files[::traj_thin]
for i,f in enumerate(files_of_interest):
print('{0}/{1}'.format(i,len(files_of_interest)))
traj = md.load(f)
distances,_ = md.compute_contacts(traj,contacts=respairs_that_changed,scheme=scheme)
X.append(distances)
print("Initial dimensionality: {0}".format(X[0].shape[1]))
print("# frames: {0}".format(np.vstack(X).shape[0]))
##### B. KINETIC DISTANCE LEARNING #####
tica = pyemma.coordinates.tica(X)
Y = tica.get_output()
# save tica model and output
np.savez_compressed('{0}_tica.npz'.format(model_name),*Y)
print("Dimensionality after tICA, retaining enough eigenvectors to explain 0.95 of kinetic variation: {0}".format(np.vstack(Y).shape[1]))
from __future__ import print_function
import mdtraj as md
import numpy as np
import itertools
t=md.load('1yrc_added.pdb') # reference
# Extract interface index
group1 = range(0,85)
group2 = range(85,98)
pairs = list(itertools.product(group1, group2))
A=md.compute_contacts(t, pairs, scheme='closest-heavy')
H=[]
A1=min(A[0]) # distance array
A2=A[1] # residue pairs
for x in range(len(A1)):
if A1[x]<=1.0: # condition 1
H.append(x)
# Extract the index from A2 which satisfy condition1 and save it to A3
A3=A2[H]
A4=[]
A5=[]
# Extract the index of residues which belong to protein
for x in range(len(A3)):
A4.append(A3[x][0])
A5=list(set(A4)) # the protein residues' index which consist of the interface
print " Loading BeadBead.dat"
beadbead = np.loadtxt("BeadBead.dat",dtype=str)
sigij = beadbead[:,5].astype(float)
epsij = beadbead[:,6].astype(float)
deltaij = beadbead[:,7].astype(float)
interaction_numbers = beadbead[:,4].astype(int)
pairs = beadbead[:,:2].astype(int)
pairs -= np.ones(pairs.shape,int)
pairs = pairs[ interaction_numbers != 0 ]
sigij = sigij[ interaction_numbers != 0 ]
print " Computing distances with mdtraj..."
traj = md.load("traj.xtc",top="Native.pdb")
distances = md.compute_contacts(traj,pairs)
contacts = (distances[0][:] <= 1.2*sigij).astype(int)
keep_frames = (((x > bounds[0]).astype(int)*(x < bounds[1]).astype(int)) == 1)
contacts = contacts[keep_frames,:]
print " Computing contact probability..."
probability = sum(contacts.astype(float))/contacts.shape[0]
C = np.zeros(Qref.shape,float)
for k in range(len(pairs)):
C[pairs[k][0],pairs[k][1]] = probability[k]
print " Plotting..."
plt.figure()
plt.subplot(1,1,1,aspect=1)
def read_and_featurize_custom(traj_file, features_dir = None, condition=None, dihedral_types = ["phi", "psi", "chi1", "chi2"], dihedral_residues = None, contact_residues = None):
#if "23" not in traj_file and "24" not in traj_file: return
top = md.load_frame(traj_file,index = 0).topology
#atom_indices = [a.index for a in top.atoms if a.residue.resSeq != 130]
atom_indices = [a.index for a in top.atoms]
traj = md.load(traj_file, atom_indices=atom_indices)
print traj_file
#print traj
#print("loaded trajectory")
'''
a = time.time()
featurizer = DihedralFeaturizer(types = ['phi', 'psi', 'chi2'])
features = featurizer.transform(traj)
b = time.time()
#print(b-a)
print("original features has dim")
print(np.shape(features))
'''
a = time.time()
dihedral_indices = []
residue_order = []
if len(dihedral_residues) > 0:
for dihedral_type in dihedral_types:
if dihedral_type == "phi": dihedral_indices.append(phi_indices(fix_topology(top), dihedral_residues))
if dihedral_type == "psi": dihedral_indices.append(psi_indices(fix_topology(top), dihedral_residues))
if dihedral_type == "chi1": dihedral_indices.append(chi1_indices(fix_topology(top), dihedral_residues))
if dihedral_type == "chi2": dihedral_indices.append(chi2_indices(fix_topology(top), dihedral_residues))
#print("new features has dim %d" %(2*len(phi_tuples) + 2*len(psi_tuples) + 2*len(chi2_tuples)))
#print("feauturizing manually:")
dihedral_angles = []
for dihedral_type in dihedral_indices:
angles = np.transpose(ManualDihedral.compute_dihedrals(traj=traj,indices=dihedral_type))
dihedral_angles.append(np.sin(angles))
dihedral_angles.append(np.cos(angles))
manual_features = np.transpose(np.concatenate(dihedral_angles))
if len(contact_residues) > 0:
fixed_traj = fix_traj(traj)
fixed_top = fixed_traj.topology
distance_residues = []
res_objects = [r for r in fixed_top.residues]
for r in contact_residues:
for res in res_objects:
if res.resSeq == r and len(res._atoms) > 5:
#print res._atoms
distance_residues.append(res.index)
if len(contact_residues) != len(distance_residues):
print "Residues are missing"
print len(contact_residues)
print len(distance_residues)
#sys.exit()
#return None
combinations = itertools.combinations(distance_residues, 2)
pairs = [c for c in combinations]
#print pairs
contact_features = md.compute_contacts(traj, contacts = pairs, scheme = 'closest-heavy', ignore_nonprotein=False)[0]
#print contact_features
#print(np.shape(contact_features))
if len(dihedral_residues) > 0:
manual_features = np.column_stack((manual_features, contact_features))
else:
manual_features = contact_features
b = time.time()
print("new features %s has shape: " %traj_file)
print(np.shape(manual_features))
if condition is None:
condition = get_condition(traj_file)
verbosedump(manual_features, "%s/%s.h5" %(features_dir, condition))
def create_features(ref, prot, lig, d):
set1 = [ref.topology.atom(i).residue.index for i in prot]
set2 = [ref.topology.atom(i).residue.index for i in lig]
contacts = md.compute_contacts(ref,contacts=list(itertools.product(set1,set2)))
atom_set = contacts[1][np.where(contacts[0]<d)[1],:]
return atom_set
eps[i][0] = a-33
elif 264<a<342:
eps[i][0] = a-67
if 29<b<176:
eps[i][1] = b-30
elif 178<b<231:
eps[i][1] = b-33
elif 264<b<342:
eps[i][1] = b-67
t1=md.load('3SN6-R.pdb')
t2=md.load('2RH1.pdb')
eps1 = [[eps[i][0], eps[i][1]] for i in range(len(eps))]
dist1=md.compute_contacts(t1, contacts=eps1, scheme='closest')
dist2=md.compute_contacts(t2, contacts=eps1, scheme='closest')
deltaDist = [dist1[0][0][i]-dist2[0][0][i] for i in range(len(dist1[0][0]))]
x = [eps[i][2] for i in range(len(eps))]
plt.scatter(x , np.absolute(deltaDist))
plt.savefig('fig1.png')
plt.show()
####################################################################################################################
# calculating dihedrals
####################################################################################################################
top1 = md.load('3SN6-R.pdb').topology
top2 = md.load('2RH1.pdb').topology
dhdrls101 = []
dhdrls102 = []
def calculate_metrics(traj, features, d):
contacts = md.compute_contacts(traj, contacts = features)
h = np.sum(contacts[0] < .5, axis=1)
return h
import mdtraj as md
import matplotlib.pyplot as plt
import numpy as np
from msmbuilder import dataset
import seaborn as sns
sns.set_style("whitegrid")
sns.set_context("poster")
#Load trajectory with ensembler models
t_models = md.load("../ensembler-models/traj-refine_implicit_md.xtc", top = "../ensembler-models/topol-renumbered-implicit.pdb")
#define 'difference' as hydrogen bond distance
k295e310 = md.compute_contacts(t_models, [[28,43]])
e310r409 = md.compute_contacts(t_models, [[43,142]])
difference = e310r409[0] - k295e310[0]
#define 'rmsd' as RMSD of activation loop from 2SRC structure
SRC2 = md.load("../reference-structures/SRC_2SRC_A.pdb")
Activation_Loop_SRC2 = [atom.index for atom in SRC2.topology.atoms if (138 <= atom.residue.index <= 158)]
Activation_Loop_Src = [atom.index for atom in t_models.topology.atoms if (138 <= atom.residue.index <= 158)]
SRC2.atom_slice(Activation_Loop_SRC2)
t_models.atom_slice(Activation_Loop_Src)
difference = difference[:,0]
