def test_average_structure(get_fn):
traj = md.load(get_fn('frame0.dcd'), top=get_fn('frame0.pdb'))
average = compute_average_structure(traj.xyz)
# The mean RMSD to the average structure should be less than to any individual frame.
sum1 = 0
sum2 = 0
for i in range(traj.n_frames):
sum1 += rmsd_qcp(traj.xyz[0], traj.xyz[i])
sum2 += rmsd_qcp(average, traj.xyz[i])
assert sum2 < sum1
def test_average_structure():
traj = md.load(get_fn('frame0.dcd'), top=get_fn('frame0.pdb'))
average = compute_average_structure(traj.xyz)
# The mean RMSD to the average structure should be less than to any individual frame.
sum1 = 0
sum2 = 0
for i in range(traj.n_frames):
sum1 += rmsd_qcp(traj.xyz[0], traj.xyz[i])
sum2 += rmsd_qcp(average, traj.xyz[i])
assert sum2 < sum1
def test_trajectory_rmsd():
t = md.load(get_fn('traj.h5'))
for parallel in [True, False]:
calculated = md.rmsd(t, t, 0, parallel=parallel)
reference = np.zeros(t.n_frames)
for i in range(t.n_frames):
reference[i] = rmsd_qcp(t.xyz[0], t.xyz[i])
eq(calculated, reference, decimal=3)
def test_trajectory_rmsd(get_fn):
t = md.load(get_fn('traj.h5'))
for parallel in [True, False]:
calculated = md.rmsd(t, t, 0, parallel=parallel)
reference = np.zeros(t.n_frames)
for i in range(t.n_frames):
reference[i] = rmsd_qcp(t.xyz[0], t.xyz[i])
eq(calculated, reference, decimal=3)
def test_precentered_1():
# test rmsd against the numpy version, using the same trajectory
# as target and reference
t1 = md.load(get_fn('traj.h5'), stride=10)
t2 = md.load(get_fn('traj.h5'), stride=10)
# don't center t1, and use it without precentered
# explicitly center t2, and use *with* precentered
for parallel in [True, False]:
t2.center_coordinates()
eq(t1.n_frames, t2.n_frames)
for i in range(t1.n_frames):
ref = np.zeros(t1.n_frames)
for j in range(t1.n_frames):
ref[j] = rmsd_qcp(t1.xyz[j], t1.xyz[i])
val1 = md.rmsd(t1, t1, i, parallel=parallel, precentered=False)
val2 = md.rmsd(t2, t2, i, parallel=parallel, precentered=True)
eq(ref, val1, decimal=3)
eq(val1, val2)
def test_precentered_1(get_fn):
# test rmsd against the numpy version, using the same trajectory
# as target and reference
t1 = md.load(get_fn('traj.h5'), stride=10)
t2 = md.load(get_fn('traj.h5'), stride=10)
# don't center t1, and use it without precentered
# explicitly center t2, and use *with* precentered
for parallel in [True, False]:
t2.center_coordinates()
eq(t1.n_frames, t2.n_frames)
for i in range(t1.n_frames):
ref = np.zeros(t1.n_frames)
for j in range(t1.n_frames):
ref[j] = rmsd_qcp(t1.xyz[j], t1.xyz[i])
val1 = md.rmsd(t1, t1, i, parallel=parallel, precentered=False)
val2 = md.rmsd(t2, t2, i, parallel=parallel, precentered=True)
eq(ref, val1, decimal=3)
eq(val1, val2)
def test_precentered_2(get_fn):
# test rmsd against the numpy version, using the difference
# trajectories as target and reference
t1_a = md.load(get_fn('traj.h5'), stride=10)
t2_a = md.load(get_fn('traj.h5'), stride=10)
t1_b = md.load(get_fn('traj.h5'), stride=10)
t2_b = md.load(get_fn('traj.h5'), stride=10)
# don't center t1, and use it without precentered
# explicitly center t2, and use *with* precentered
t2_a.center_coordinates()
t2_b.center_coordinates()
for parallel in [True, False]:
for i in range(t1_b.n_frames):
ref = np.zeros(t1_a.n_frames)
for j in range(t1_a.n_frames):
ref[j] = rmsd_qcp(t1_a.xyz[j], t1_b.xyz[i])
val1 = md.rmsd(t1_a, t1_b, i, parallel=parallel, precentered=False)
val2 = md.rmsd(t2_a, t2_b, i, parallel=parallel, precentered=True)
eq(ref, val1, decimal=3)
eq(val1, val2, decimal=4)
def test_precentered_2():
# test rmsd against the numpy version, using the difference
# trajectories as target and reference
t1_a = md.load(get_fn('traj.h5'), stride=10)
t2_a = md.load(get_fn('traj.h5'), stride=10)
t1_b = md.load(get_fn('traj.h5'), stride=10)
t2_b = md.load(get_fn('traj.h5'), stride=10)
# don't center t1, and use it without precentered
# explicitly center t2, and use *with* precentered
t2_a.center_coordinates()
t2_b.center_coordinates()
for parallel in [True, False]:
for i in range(t1_b.n_frames):
ref = np.zeros(t1_a.n_frames)
for j in range(t1_a.n_frames):
ref[j] = rmsd_qcp(t1_a.xyz[j], t1_b.xyz[i])
val1 = md.rmsd(t1_a, t1_b, i, parallel=parallel, precentered=False)
val2 = md.rmsd(t2_a, t2_b, i, parallel=parallel, precentered=True)
eq(ref, val1, decimal=3)
eq(val1, val2, decimal=4)
# But for some applications like clustering, we want to run many
# rmsd() calculations per trajectory frame. Under these circumstances,
# the centering of the trajectories is going to be done many times, which
# leads to a slight slowdown. If we manually center the trajectory
# and then inform the rmsd() function that the centering has been
# precomputed, we can achieve ~2x speedup, depending on your machine
# and the number of atoms.
t.center_coordinates()
start = time.time()
for i in range(100):
md.rmsd(t, t, 0, precomputed=True)
print 'md.rmsd(precomputed=True): %.2f rmsds / s' % ((t.n_frames * 100) / (time.time() - start))
# Just for fun, let's compare this code to the straightforward
# numpy implementation of the same algorithm, which mdtraj has
# (mostly for testing) in the mdtraj.geometry.alignment subpackage
from mdtraj.geometry.alignment import rmsd_qcp
start = time.time()
for k in range(t.n_frames):
rmsd_qcp(t.xyz[0], t.xyz[k])
print 'pure numpy rmsd_qcp(): %.2f rmsds / s' % (t.n_frames / (time.time() - start))
# The :func:`md.rmsd()` code is *a lot* faster. If you go look at the :func:`rmsd_qcp`
# source code, you'll see that it's not because that code is particularly slow or
# unoptimized. It's about as good as you can do with numpy. The reason for the speed
# difference is that an inordinate amount of time was put into hand-optimizing
# an SSE3 implementation in C for the :func:`md.rmsd()` code.
def test_rmsd_nonzero():
rmsd_kabsch = alignment.rmsd_kabsch(xyz1, xyz3)
rmsd_qcp = alignment.rmsd_qcp(xyz1, xyz3)
eq(rmsd_kabsch, rmsd_qcp, decimal=5)
def test_rmsd_zero():
rmsd_kabsch = alignment.rmsd_kabsch(xyz1, xyz2)
rmsd_qcp = alignment.rmsd_qcp(xyz1, xyz2)
eq(float(rmsd_kabsch), 0.0, decimal=5)
eq(float(rmsd_qcp), 0.0, decimal=5)
def test_rmsd_nonzero():
rmsd_kabsch = alignment.rmsd_kabsch(xyz1, xyz3)
rmsd_qcp = alignment.rmsd_qcp(xyz1, xyz3)
eq(rmsd_kabsch, rmsd_qcp, decimal=5)
def test_rmsd_zero():
rmsd_kabsch = alignment.rmsd_kabsch(xyz1, xyz2)
rmsd_qcp = alignment.rmsd_qcp(xyz1, xyz2)
eq(float(rmsd_kabsch), 0.0, decimal=5)
eq(float(rmsd_qcp), 0.0, decimal=5)