def test_2_scalar_couplings(get_fn):
t = md.load(get_fn('frame0.h5')) # This is Alanine dipeptide
for model in ["Ruterjans1999", "Bax2007", "Bax1997"]:
indices, J = md.compute_J3_HN_HA(t)
eq(indices.shape, (1, 4))
eq(J.shape, (501, 1))
J = J.mean()
assert abs(J - 6.06) <= 2.0, "Value is far from experimental value."
def test_2_scalar_couplings():
t = md.load(get_fn('frame0.h5')) # This is Alanine dipeptide
for model in ["Ruterjans1999", "Bax2007", "Bax1997"]:
indices, J = md.compute_J3_HN_HA(t)
eq(indices.shape, (1, 4))
eq(J.shape, (501, 1))
J = J.mean()
assert abs(J - 6.06) <= 2.0, "Value is far from experimental value."
def test_3_scalar_couplings(get_fn):
t = md.load(get_fn('1bpi.pdb'))
for model in ["Bax2007"]:
indices_HA, J_HA = md.compute_J3_HN_HA(t)
indices_C, J_C = md.compute_J3_HN_C(t)
indices_CB, J_CB = md.compute_J3_HN_CB(t)
eq(indices_HA.shape, (57, 4))
eq(indices_C.shape, (57, 4))
eq(indices_CB.shape, (57, 4))
eq(J_HA.shape, (1, 57))
eq(J_C.shape, (1, 57))
eq(J_CB.shape, (1, 57))
np.testing.assert_almost_equal(J_HA[0, 0], 0.48885268)
np.testing.assert_almost_equal(J_C[0, 0], 3.840529)
np.testing.assert_almost_equal(J_CB[0, 0], 2.5702963)
def test_3_scalar_couplings(get_fn):
t = md.load(get_fn('1bpi.pdb'))
for model in ["Bax2007"]:
indices_HA, J_HA = md.compute_J3_HN_HA(t)
indices_C, J_C = md.compute_J3_HN_C(t)
indices_CB, J_CB = md.compute_J3_HN_CB(t)
eq(indices_HA.shape, (57, 4))
eq(indices_C.shape, (57, 4))
eq(indices_CB.shape, (57, 4))
eq(J_HA.shape, (1, 57))
eq(J_C.shape, (1, 57))
eq(J_CB.shape, (1, 57))
np.testing.assert_almost_equal(J_HA[0, 0], 0.48885268)
np.testing.assert_almost_equal(J_C[0, 0], 3.840529)
np.testing.assert_almost_equal(J_CB[0, 0], 2.5702963)
def analyze(self, traj):
top, bonds = traj.top.to_dataframe()
ind, values = md.compute_J3_HN_HA(traj)
prediction = pd.DataFrame({"value":values.mean(0)})
prediction["resSeq"] = top.ix[ind[:, -1]].resSeq.values # Set the residue numbers to the last (fourth) atom in the dihedral
if top.ix[0].resName == "ACE":
prediction["resSeq"] -= 1 # HARDCODED Hack to account for the ACE residue!!!!!!!!!! Fix me later!
prediction["AA"] = top.ix[ind[:, -1]].resName.values
prediction["expt"] = "3JHNHA"
prediction["system"] = self.identifier
prediction["sigma"] = 0.36
multi_index = prediction.set_index(["system", "expt", "resSeq"]).index
prediction["identifier"] = multi_index_to_str(multi_index)
prediction = prediction.set_index("identifier")
return prediction
def analyze(self, traj):
top, bonds = traj.top.to_dataframe()
ind, values = md.compute_J3_HN_HA(traj)
prediction = pd.DataFrame({"value": values.mean(0)})
prediction["resSeq"] = top.ix[
ind[:,
-1]].resSeq.values # Set the residue numbers to the last (fourth) atom in the dihedral
if top.ix[0].resName == "ACE":
prediction[
"resSeq"] -= 1 # HARDCODED Hack to account for the ACE residue!!!!!!!!!! Fix me later!
prediction["AA"] = top.ix[ind[:, -1]].resName.values
prediction["expt"] = "3JHNHA"
prediction["system"] = self.identifier
prediction["sigma"] = 0.36
multi_index = prediction.set_index(["system", "expt", "resSeq"]).index
prediction["identifier"] = multi_index_to_str(multi_index)
prediction = prediction.set_index("identifier")
return prediction
def index2res(structure):
"""
The function creates list of residues based on files with indexes, used to calculate
J3 coupling constants. This stuff should have been done during J3 calculations, because
outputing atom indexes is meaningless, as they depend on topology.For example, amide
nitrogen of the second residue may have different atomic number in pdb, containing only
backbone,than in pdb, containing both backbone and sidechain.
In the input file, each line correspond to 4 indexes of atoms, which form a dihedral
angle, used to compute J3. For Halpha-HN, this list includs atoms, forming phi
dihedral angle. These are C(i-1),N(i),Ca(i),C(i). We are interested in residue,
Nitrogen from which participates in forming a bond. So we are looling for the 2 atom
(1 by Python numeration)
Input:
structure - mdtraj frame
Output:
residue_index - list of Python-numerated residue ids.
residue_name
"""
index, J3_inp = md.compute_J3_HN_HA(structure)
assert (index.shape[0] == structure.n_residues - 1)
residue_index = []
residue_name = []
for i in range(0, index.shape[0]):
atom_index = int(index[i, 1])
assert (structure.top.atom(atom_index).name == 'N')
residue_index.append(structure.top.atom(atom_index).residue.index)
residue_name.append(structure.top.atom(atom_index).residue.name)
return (residue_index, residue_name)
def test_3_scalar_couplings():
t = md.load(get_fn('1bpi.pdb'))
for model in ["Ruterjans1999", "Bax2007", "Bax1997"]:
indices, J = md.compute_J3_HN_HA(t)
def test_3_scalar_couplings():
t = md.load(get_fn('1bpi.pdb'))
for model in ["Ruterjans1999", "Bax2007", "Bax1997"]:
indices, J = md.compute_J3_HN_HA(t)
metavar='pairs',
type=str,
help='filename of pairs to calc 3J const')
parser.add_argument('weight',
metavar='weight',
type=str,
help='filename of weights')
args = parser.parse_args()
xtc = args.xtc
pdb = args.pdb
basename = args.basename
t = md.load(xtc, top=pdb)
list, J3_HN_HA = md.compute_J3_HN_HA(t)
pairs = []
f = open(args.pairs, 'r')
l = f.readlines()
f.close()
for item in l:
pairs.append(item)
w = []
f = open(args.weight, 'r')
l = f.readlines()
f.close()
for item in l:
i = item.split()
w.append(float(i[1]))
parsed = nmrpystar.parse(open("./19127.str").read())
print(parsed.status)
q = parsed.value.saves["coupling_constant_list_1"].loops[1]
x = pd.DataFrame(q.rows, columns=q.keys)
x = x[[
"Coupling_constant.Seq_ID_1", "Coupling_constant.Val",
"Coupling_constant.Val_err"
]]
x.rename(columns={
"Coupling_constant.Seq_ID_1": "resSeq",
"Coupling_constant.Val": "value",
"Coupling_constant.Val_err": "err"
},
inplace=True)
# Need to make dtypes match to do eventual comparison.
x["resSeq"] = x["resSeq"].astype('int')
x["value"] = x["value"].astype('float')
expt = x.set_index(["resSeq"]).value
top, bonds = t.top.to_dataframe()
ind, values = md.compute_J3_HN_HA(t)
prediction = pd.Series(values.mean(0), top.ix[ind[:, -1]].resSeq)
delta = (expt - prediction).dropna()
delta.name = "value"
rms = (delta**2.).mean(0)**0.5
parsed = nmrpystar.parse(open("./19127.str").read())
print(parsed.status)
q = parsed.value.saves["coupling_constant_list_1"].loops[1]
x = pd.DataFrame(q.rows, columns=q.keys)
x = x[["Coupling_constant.Seq_ID_1", "Coupling_constant.Val", "Coupling_constant.Val_err"]]
x.rename(
columns={
"Coupling_constant.Seq_ID_1": "resSeq",
"Coupling_constant.Val": "value",
"Coupling_constant.Val_err": "err",
},
inplace=True,
)
# Need to make dtypes match to do eventual comparison.
x["resSeq"] = x["resSeq"].astype("int")
x["value"] = x["value"].astype("float")
expt = x.set_index(["resSeq"]).value
top, bonds = t.top.to_dataframe()
ind, values = md.compute_J3_HN_HA(t)
prediction = pd.Series(values.mean(0), top.ix[ind[:, -1]].resSeq)
delta = (expt - prediction).dropna()
delta.name = "value"
rms = (delta ** 2.0).mean(0) ** 0.5
reference.append(xyz)
reference = pd.DataFrame(reference, columns=["seq", "resSeq", "value"])
reference = reference.set_index(["seq", "resSeq"]).value
reference = reference.drop_duplicates()
data = []
for (ff, water, seq) in products:
try:
aa0, aa1 = seq.split("_")[1]
aa_string = "%s%s" % (aa0, aa1)
t = md.load("./dcd/%s_%s_%s.dcd" % (ff, water, seq), top="./pdbs/%s.pdb" % (seq))[1500:]
except:
continue
#phi = md.compute_phi(t)[1] * 180 / np.pi
#J0, J1 = scalar_couplings.J3_HN_HA(phi).mean(0)
J0, J1 = md.compute_J3_HN_HA(t)[1].mean(0)
data.append([ff, water, aa_string, 0, J0])
data.append([ff, water, aa_string, 1, J1])
data = pd.DataFrame(data, columns=["ff", "water", "seq", "resSeq", "value"])
X = data.pivot_table(cols=["seq", "resSeq"], rows=["ff", "water"], values="value")
delta = X - reference
Z = (delta / 0.36)
rms_by_model = (Z ** 2.).mean(1) ** 0.5
rms_by_model
