Copy function did not return an independent entity?

[locitran]

Hi,

I try to do superimposition u2 into u1 using CA only. These structures have the same number of atoms.
After translating and rotating u1_ca, I found u1 coordinate also changed as well, even copy() functions had used.

Is this a normal situation or a bug? How could I make a complete independent entity?
Thanks

Please download the files here: u1.pdb and u2.pdb

import MDAnalysis as mda
from MDAnalysis.analysis import align

pdb1 = 'u1.pdb'
pdb2 = 'u2.pdb'
u1 = mda.Universe(pdb1)
u2 = mda.Universe(pdb2)
u1_ca = u1.select_atoms('name CA').copy()
u2_ca = u2.select_atoms('name CA').copy()
print(u1.select_atoms('name CA').positions[1:3,1:3], u1_ca.positions[1:3,1:3])

Before superimposition,

(array([[-5.735, 28.569],
        [-4.479, 26.143]], dtype=float32),
 array([[-5.735, 28.569],
        [-4.479, 26.143]], dtype=float32))

mobile0 = u1_ca.positions - u1_ca.atoms.center_of_mass()
ref0 = u2_ca.positions - u2_ca.atoms.center_of_mass()
R, rmsd_value = align.rotation_matrix(mobile0, ref0)
print("Rotation matrix = \n", R)
print("RMSD = ", rmsd_value)

u1_ca.atoms.translate(-u1_ca.atoms.center_of_mass())
u1_ca.atoms.rotate(R)
u1_ca.atoms.translate(u2_ca.atoms.center_of_mass())
print(u1.select_atoms('name CA').positions[1:3,1:3], u1_ca.positions[1:3,1:3])

After superimposition,

(array([[  9.291354, -24.866745],
        [  8.894788, -21.551838]], dtype=float32),
 array([[  9.291354, -24.866745],
        [  8.894788, -21.551838]], dtype=float32))

[orbeckst]

The line

u1_ca = u1.select_atoms('name CA').copy()

makes a copy of an AtomGroup. However, an AtomGroup is (essentially) just an array of indices into the underlying coordinates (namely u.trajectory.ts.positions). It is effectively the same as

u1_ca = u1.select_atoms('name CA')

Copying it does not give you a copy of the underlying coordinates. You can make copies of coordinates at the current frame and before the superposition with

u1_pos = u1_ca.positions.copy()

because positions is a regular numpy array. Then compare u1_pos with u1_ca.positions.

See also Why do the atom position change over trajectories?