j-wags
commented
4 years ago Hi @kcreemer,
Thanks for the thoroughly-written issue.
Are there alternative ways to fragmentate the polymer so charge calculations do not have to be performed for the molecule itself but for the units it consists of?
We're planning to develop a workflow to automatically fragment and charge polymers like this. This workflow would allow users to define a protein force field, a residue separator SMARTS, and a method for treating nonstandard amino acids and capping groups. Unfortunately, we have several more pressing features on our roadmap, so I can't say whether this will be ready within a year. In the short-term, you're probably best off defining LibraryCharges for your polymer, using the 10-mer approach that you mentioned above.
Can we define a new library residue so this can be recognized within our molecule?
Yes, and this is the approach I would suggest. Use examples for the LibraryCharges tag can be found in the SMIRNOFF spec documentation. This will require writing three SMARTS strings by hand (which can be a bit painful), but once it works, it should suffice for polymers of arbitrary length. As you mentioned, the three LibraryCharge elements that would need to be defined are the initiator, terminator, and monomer unit. If writing the SMARTS is prohibitively difficult, we provide functionality in Chemper for machine-generating SMARTS patterns, however it is not as geared for external use as the OFF Toolkit, so the installation/documentation may not be as mature.
Alternatively use: create_openmm_system(topology, charge_from_molecules=molecule_list)?
This would also work, though you'd need to manually update offmol.partial_charges with your desired charges ahead of time. I'd be cautious about this, just because the workflow would need to assign a large number of charges in exactly the right order to match up with the atom indexing. If, for any reason, the molecule atom indexing changed or was misinterpreted, the partial charges assigned to the resulting system would silently be scrambled. For this reason, I'd recommend using LibraryCharges, since it ensures that charges are assigned based on the connectivity of the model.
Let's leave this issue open until we get a solution working for you. Please let me know if there's anything else I can do to help.
kcreemer
Describe the bug To study the interaction between a polymer and an API, we want to parameterize polymers ranging from 50 to 600 monomer units. This to perform a molecular dynamics calculation with openff. However from the moment we reach +300 atoms, calculations of the partial charges is no longer possible.
To Reproduce The setup of the system is done using smiles codes and openbabel (and pybel). It is implemented via the following way: off_polymer_system = off_forcefield.create_openmm_system(off_polymer_topology)
Output Exception Traceback (most recent call last)