CA filtering in the evaluation code

[ynshen]

Hi @bjing2016,

Thanks for open-sourcing this awesome work!

I am running through the evaluation code scripts/analyze_ensembles.py but found this line might not correctly filter CA atoms for the sample ensemble aftraj.

alphaflow/scripts/analyze_ensembles.py

Line 154 in 849e7a2

out['ca_mask'] = ca_mask = [a.index for a in traj_aa.top.atoms if a.name == 'CA']

# For example: name = 6y2x_A. aftraj is the ensemble generated by AlphaFlow-MD.
>>> [a for a in traj.top.atoms]
[GLY1-CA,
 SER2-CA,
 GLU3-CA,
 PRO4-CA,
 GLU5-CA,
 PRO6-CA,
 GLU7-CA,
...]

>>> [a for a in aftraj.top.atoms]
[GLY1-CA,
 SER2-CA,
 GLU3-CA,
 PRO4-C,
 GLU5-CA,
 PRO6-C,
 GLU7-CA,
...]

It is likely due to different atom ordering in the topology files for certain amino acids. Would need to filter them separately to ensure getting the CA-only ensemble:

af_ca_mask = [a.index for a in aftraj_aa.top.atoms if a.name == 'CA']
aftraj_fixed = aftraj_aa.atom_slice(af_ca_mask, False)