Proper error message and suggestion for AlignTraj on trajectory without mass

[xiki-tempula]

Is your feature request related to a problem? Please describe.
Coarse-grain trajectory has no assigned mass but it still makes sense to use the AlignTraj. Thus, if one were to run.

u = mda.Universe(gro)
protein = u.select_atoms('bynum 1-518')
traj = mda.Universe(xtc, xtc)
AlignTraj(traj, protein, select='bynum 1-518', in_memory=True)

He will get

  File "pa.py", line 13, in plot
    AlignTraj(traj, protein, select='bynum 1-518', in_memory=True)
  File "/Users//GitHub/mdanalysis/package/MDAnalysis/analysis/align.py", line 668, in __init__
    strict=strict, match_atoms=match_atoms)
  File "/Users//GitHub/mdanalysis/package/MDAnalysis/analysis/align.py", line 1187, in get_matching_atoms
    mass_mismatches = (np.absolute(ag1.masses - ag2.masses) > tol_mass)
  File "/Users//GitHub/mdanalysis/package/MDAnalysis/core/groups.py", line 2294, in __getattr__
    cls=self.__class__.__name__, attr=attr))
AttributeError: AtomGroup has no attribute masses

I felt that there should be certain options to disable the check on mass if the weight is not set to mass.

Say if one run

AlignTraj(traj, protein, select='bynum 1-518', in_memory=True, weights=None)

Then the AlignTraj should work without checking for mass.

[richardjgowers]

Can this issue be worked around by setting all masses to 1 (as your beads happen to have identical masses)?

…

On Sun, Jan 26, 2020 at 21:37, Zhiyi Wu ***@***.***> wrote: *Is your feature request related to a problem? Please describe.* Coarse-grain trajectory has no assigned mass but it still makes sense to use the AlignTraj. Thus, if one were to run. u = mda.Universe(gro) protein = u.select_atoms('bynum 1-518') traj = mda.Universe(xtc, xtc) AlignTraj(traj, protein, select='bynum 1-518', in_memory=True) He will get File "pa.py", line 13, in plot AlignTraj(traj, protein, select='bynum 1-518', in_memory=True) File "/Users//GitHub/mdanalysis/package/MDAnalysis/analysis/align.py", line 668, in __init__ strict=strict, match_atoms=match_atoms) File "/Users//GitHub/mdanalysis/package/MDAnalysis/analysis/align.py", line 1187, in get_matching_atoms mass_mismatches = (np.absolute(ag1.masses - ag2.masses) > tol_mass) File "/Users//GitHub/mdanalysis/package/MDAnalysis/core/groups.py", line 2294, in __getattr__ cls=self.__class__.__name__, attr=attr)) AttributeError: AtomGroup has no attribute masses I felt that there should be certain options to disable the check on mass if the weight is not set to mass. Say if one run AlignTraj(traj, protein, select='bynum 1-518', in_memory=True, weights=None) Then the AlignTraj should work without checking for mass. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2469?email_source=notifications&email_token=ACGSGB72YE43NFV4RQKQDQLQ7X7CTA5CNFSM4KLZLXD2YY3PNVWWK3TUL52HS4DFUVEXG43VMWVGG33NNVSW45C7NFSM4IIZCRSQ>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACGSGB24TI7J4X6EYMWOFD3Q7X7CTANCNFSM4KLZLXDQ> .

[xiki-tempula]

@richardjgowers Yes, I'm currently solving it by adding the mass. But I felt that since the AlignTraj doesn't require mass information in this case, it doesn't make sense by force the mass attribute to be present,

[lilyminium]

Try setting match_atoms=False. The only reason it checks masses is to try to enforce that the reference and mobile atoms are in the same order.

Edit: although I think that's only in the develop branch. And possibly it should be off by default anyway.

[xiki-tempula]

@lilyminium Thanks. It seems to be working. It might be a good idea to document it.

[richardjgowers]

I think this just needs a doc fix rather than a behaviour change. The case where all atoms are similar mass (so don't required weighting) is very niche, and I can see more people (with dissimilar masses) getting burnt by weighting being ignored when masses aren't available.

[lilyminium]

@richardjgowers weights=None by default in AlignTraj, so it doesn't weight alignment without explicitly turning it on. In this case it's dying because get_matching_atoms looks for 'mismatching' atoms by checking if any one pair has a difference higher than tol_mass. If you want to keep match_atoms=True by default then you could just change this line

mdanalysis/package/MDAnalysis/analysis/align.py

Line 1367 in 3be5c98

if match_atoms:

to if match_atoms and hasattr(ag1, 'masses') and hasattr(ag2, 'masses')

[siddharthjain1611]

@lilyminium I'm making the changes suggested by you hope these try to solve the problem of mismatching.

[siddharthjain1611]

@lilyminium If could please spare some time to review my PR and suggest the changes required.

[lilyminium]

@xiki-tempula actually I'm surprised your topology didn't have masses, because I thought that was a mandatory attribute that would be guessed if not given.