def analyze_pca(u: mda.Universe, n_dimensions=40):
"""Fetch PCA component contribution values for a single trajectory."""
pca_analysis = pca.PCA(u, select='backbone')
space = pca_analysis.run()
space_3 = space.transform(u.select_atoms('backbone'), 3)
w = pca.cosine_content(space_3, 0)
print(w)
return [
space.variance[:n_dimensions], space.cumulated_variance[:n_dimensions]
]
def test_cosine_content():
rand = MDAnalysis.Universe(RANDOM_WALK_TOPO, RANDOM_WALK)
pca_random = pca.PCA(rand).run()
dot = pca_random.transform(rand.atoms)
content = pca.cosine_content(dot, 0)
assert_almost_equal(content, .99, 1)
def test_cosine_content():
rand = mda.Universe(RANDOM_WALK_TOPO, RANDOM_WALK)
pca_random = PCA(rand).run()
dot = pca_random.transform(rand.atoms)
content = cosine_content(dot, 0)
assert_almost_equal(content, .99, 1)
def main():
"""Run main procedure."""
# TODO(schneiderfelipe): accept multiple files
parser = argparse.ArgumentParser(description=__doc__)
parser.add_argument("traj_files", nargs="+")
args = parser.parse_args()
gnorms = []
energies = []
all_atomnos = []
all_atomcoords = []
for traj_file in args.traj_files:
atomnos, comments, atomcoords = read_xyz(traj_file)
all_atomnos.extend(atomnos)
all_atomcoords.extend(atomcoords)
for comment in comments:
fields = comment.split()
gnorms.append(float(fields[3]))
energies.append(float(fields[1]))
energies = np.array(energies)
energies -= energies.min()
energies *= hartree * N_A / (kilo * calorie)
u = mda.Universe.empty(n_atoms=len(all_atomnos[0]), trajectory=True)
u.add_TopologyAttr("type", [element[i] for i in all_atomnos[0]])
u.load_new(all_atomcoords, order="fac")
print(u)
selection = None
print("(enter 'q' for exit, 'h' for help)")
while True:
code = input("select> ").strip().split()
if code[0] == "q":
break
elif code[0] == "h":
for key in commands:
print(f"{key:15s}: {commands[key]}")
elif code[0] == "e":
fig, ax = plt.subplots(2)
ax[0].plot(energies)
ax[0].set_xlabel("frame")
ax[0].set_ylabel("energy (kcal/mol)")
ax[1].plot(gnorms)
ax[1].set_xlabel("frame")
ax[1].set_ylabel("grad. norm (Eh/a0)")
plt.show()
elif code[0] == "s":
print(selection)
if selection is not None:
print(selection_text)
elif code[0] == "pca":
if selection is None:
print("empty selection, doing nothing")
continue
p = pca.PCA(u, select=selection_text)
p.run()
n_pcs = np.where(p.cumulated_variance > 0.95)[0][0]
print(n_pcs)
print(p.cumulated_variance[0:n_pcs])
pca_space = p.transform(selection, n_components=n_pcs)
print(pca_space)
print(pca.cosine_content(pca_space, 0))
elif code[0] == "p":
if selection is None:
print("empty selection, doing nothing")
continue
n = len(selection)
if n == 2:
data_label = "bond length (Å)"
elif n == 3:
data_label = "bond angle (°)"
elif n == 4:
data_label = "dihedral angle (°)"
else:
print("too few or too many indices")
continue
data = []
for i, (e, ts) in enumerate(zip(energies, u.trajectory)):
if n == 2:
d = distances.calc_bonds(
selection[0].position, selection[1].position
)
elif n == 3:
d = np.degrees(
distances.calc_angles(
selection[0].position,
selection[1].position,
selection[2].position,
)
)
elif n == 4:
d = np.degrees(
distances.calc_dihedrals(
selection[0].position,
selection[1].position,
selection[2].position,
selection[3].position,
)
)
data.append(d)
if i % 100 == 0 or i == len(u.trajectory) - 1:
print(
f"frame = {ts.frame:4d}: e = {e:5.1f} kcal/mol, {data_label.split('(')[0][:-1]} = {d:7.3f} {data_label[-2]}"
)
data = np.array(data)
fig, ax = plt.subplots(1, 2)
ax[0].plot(data)
ax[0].set_xlabel("frame")
ax[0].set_ylabel(data_label)
ax[1].plot(energies, data, "o", label="data points")
ax[1].set_xlabel("energy (kcal/mol)")
ax[1].set_ylabel(data_label)
if n == 2:
dx = 0.1
elif n == 3:
dx = 10.0
elif n == 4:
dx = 10.0
res = stats.binned_statistic(
data, energies, "min", min(25, (data.max() - data.min()) / dx)
)
# print(res.statistic)
mask = np.isnan(res.statistic)
res.statistic[mask] = np.interp(
np.flatnonzero(mask), np.flatnonzero(~mask), res.statistic[~mask]
)
# print(res.statistic)
# ax[1].hlines(res.statistic, res.bin_edges[:-1], res.bin_edges[1:], colors='g', lw=2, label='binned min. energies')
ax[1].barh(
(res.bin_edges[:-1] + res.bin_edges[1:]) / 2,
res.statistic,
align="center",
height=res.bin_edges[1:] - res.bin_edges[:-1],
alpha=0.25,
label="binned min. energies",
)
ax[1].legend()
plt.show()
else:
try:
selection_text = " ".join(code)
selection = u.select_atoms(selection_text)
except mda.exceptions.SelectionError as e:
print(e)
print("bye")