Unable to allocate memory for an array with shape

[jochuan]

I am trying to martinize a protein complex in a single run to generate Go-Martini contacts between the two proteins, but I keep getting an “out of memory” error even though my system still shows 23.6 GiB of free RAM. Is there anything I can do to work around this issue, or do I simply need to wait for the changes in pull request #733 to be merged?

[pckroon]

You could try running the version in that branch :)
As far as I can see the only open issue with the PR is that the tests need to be adjusted to account for the fact that the arrays can be sparse.

[jochuan]

I don't know if i made everything right but i am getting a crash when generating the contact map with that PR

[pckroon]

How did you install and run? Did you get a more specific error?

[jochuan]

git clone https://github.com/marrink-lab/vermouth-martinize.git
cd vermouth-martinize
git fetch origin pull/733/head:pr733
git checkout pr733
pip install .

My command was
martinize2 -f model_00_CAT_GOX_protonated.pdb -o topol.top -x model_00_CAT_GOX_CG.pdb -ff martini3001 -dssp -go -p backbone

And the crash is the following:

INFO - general - Identified the modifications ['LYS-HZ3'] on residues ['LYS231']
    INFO - general - Identified the modifications ['LYS-HZ3'] on residues ['LYS297']
    INFO - general - Identified the modifications ['C-ter'] on residues ['LEU366', 'LEU366', 'LEU366']
    INFO - step - Read input.
    INFO - step - Generating Go model contact map.
Traceback (most recent call last):
  File "/home/jochuan/anaconda3/lib/python3.11/site-packages/vermouth/rcsu/contact_map.py", line 317, in _get_vdw_radius
    atom_vdw = res_vdw[atomname]
               ~~~~~~~^^^^^^^^^^
KeyError: 'N'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/jochuan/anaconda3/bin/martinize2", line 1205, in <module>
    entry()
  File "/home/jochuan/anaconda3/bin/martinize2", line 1017, in entry
    GenerateContactMap(write_file=args.go_write_file).run_system(system)
  File "/home/jochuan/anaconda3/lib/python3.11/site-packages/vermouth/rcsu/contact_map.py", line 840, in run_system
    contacts = self.run_molecule(molecule)
               ^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jochuan/anaconda3/lib/python3.11/site-packages/vermouth/rcsu/contact_map.py", line 836, in run_molecule
    return do_contacts(molecule, self.write_file)
           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jochuan/anaconda3/lib/python3.11/site-packages/vermouth/rcsu/contact_map.py", line 798, in do_contacts
    vdw_list, atypes, coords, res_serial, resids, chains, resnames, res_idx, ca_pos, nresidues, mol_graph = _contact_info(
                                                                                                            ^^^^^^^^^^^^^^
  File "/home/jochuan/anaconda3/lib/python3.11/site-packages/vermouth/rcsu/contact_map.py", line 448, in _contact_info
    vdw_list.append(_get_vdw_radius(subgraph.nodes[atom]['resname'],
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/jochuan/anaconda3/lib/python3.11/site-packages/vermouth/rcsu/contact_map.py", line 319, in _get_vdw_radius
    atom_vdw = res_vdw['default']
               ~~~~~~~^^^^^^^^^^^

Sorry if i made something wrong but i don't have much knowledge in coding and using github

[csbrasnett]

That error suggests to me that you have non-standard residues in your protein, which don't have associated data for the Go model. @pckroon I think that PR (or another, I can pick up) needs some better error handling for this kind of thing, and I've left a comment for some now redundant code

[jochuan]

The structure that i am trying to martinize was generated by the CHARMM-GUI PDB Reader & Manipulator. And this crash doesn't happen when using the normal version of martinize2 without the PR merged and martinizing one protein at a time instead of the whole complex in one go

[csbrasnett]

Yes, that makes sense, because the error is coming from the contact map calculator. I suspect the residue in question is some kind of HIS residue where these parameters haven't been written explicitly. Otherwise, does your structure have glycosylated/PTM etc residues? That would certainly cause an issue along these lines. If it is the HIS issue, this can be fixed, but it may not be immediate apologies